Measured browser compatibility for LiteRT (.tflite) models: what loads, what delegates to WebGPU, what matches CPU outputs, and at what latency. Passing and failing models are listed — a model that does not run in the browser is information.
Every number is labeled with the environment that produced it — headless desktop Chromium is not a user’s phone, and results are only meaningful together with their environment. Click a column header to sort.
| Model | Task | Native delegation | wasm_xnnpack | webgpu_mldrift | Device runs | Card | Demo |
|---|---|---|---|---|---|---|---|
3ddfa-v2 | keypoint-detection | — | pass · 25.31 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | pass · 1.255 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | raspberry-pi-5 cpu_xnnpack: pass · 3.539 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
6drepnet-headpose | image-classification | — | pass · 240.73 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | pass · 17.325 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | raspberry-pi-5 cpu_xnnpack: pass · 140.556 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
audio8-tts-preview-0.6b__codec_decoder_fp16_t128 | text-to-speech | gpu_mldrift @ litert 2.2.0: 81.6% · 154 partition(s) | run_failed stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-09-28 | load_failed stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-09-28 | galaxy-s26 cpu_xnnpack: fallback · 11109.046 msGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.2 · Android 16 (build BP4A.251205.006.S942QOPS1AZH9) · 2026-11-02galaxy-s26 gpu: output_mismatch stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (build BP4A.251205.006.S942QOPS1AZH9) · 2026-10-10galaxy-s26 gpu_mldrift: load_failed stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (build BP4A.251205.006.S942QOPS1AZH9) · 2026-11-01galaxy-s26 gpu_mldrift: output_mismatch · 962.25 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (build BP4A.251205.006.S942QOPS1AZH9) · 2026-11-01galaxy-s26 gpu_mldrift: output_mismatch · 2460.437 msGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.2 · Android 16 (build BP4A.251205.006.S942QOPS1AZH9) · 2026-11-02galaxy-s26 gpu_mldrift: output_mismatch · 1052.356 msGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.2 · Android 16 (build BP4A.251205.006.S942QOPS1AZH9) · 2026-11-01galaxy-s26 gpu_mldrift: output_mismatch stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (build BP4A.251205.006.S942QOPS1AZH9) · 2026-11-02galaxy-s26 gpu_mldrift: output_mismatch · 1053.005 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (build BP4A.251205.006.S942QOPS1AZH9) · 2026-11-01galaxy-s26 npu_qnn: run_failed stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · QAIRT 2.47.0.260601114230 HTP libraries with the LiteRT 2.2.0 Qualcomm compiler plugin and dispatch library (all MATCH) · Android 16 (build BP4A.251205.006.S942QOPS1AZH9) · 2026-11-01pixel-11-pro cpu_xnnpack: fallback · 10257.055 msPixel 11 Pro · Google Tensor G6 · litert 2.3.2 · Android 17 (build CD1A.260618.001.A7) · 2026-11-02pixel-11-pro gpu_mldrift: load_failed stale — latest known 2.3.0Pixel 11 Pro · Google Tensor G6 · litert 2.2.0 · Android 17 (build CD1A.260618.001.A7) · 2026-11-01pixel-11-pro gpu_mldrift: output_mismatch · 1157.462 ms stale — latest known 2.3.0Pixel 11 Pro · Google Tensor G6 · litert 2.2.0 · Android 17 (build CD1A.260618.001.A7) · 2026-11-01pixel-11-pro gpu_mldrift: output_mismatch · 2688.813 msPixel 11 Pro · Google Tensor G6 · litert 2.3.2 · Android 17 (build CD1A.260618.001.A7) · 2026-11-01pixel-11-pro gpu_mldrift: output_mismatch · 1171.223 msPixel 11 Pro · Google Tensor G6 · litert 2.3.2 · Android 17 (build CD1A.260618.001.A7) · 2026-11-01pixel-11-pro gpu_mldrift: output_mismatch stale — latest known 2.3.0Pixel 11 Pro · Google Tensor G6 · litert 2.2.0 · Android 17 (build CD1A.260618.001.A7) · 2026-11-02pixel-11-pro gpu_mldrift: output_mismatch · 1280.521 ms stale — latest known 2.3.0Pixel 11 Pro · Google Tensor G6 · litert 2.2.0 · Android 17 (build CD1A.260618.001.A7) · 2026-11-01 | CARD.md · card.json | — |
audio8-tts-preview-0.6b__codec_decoder_fp16_t192 | text-to-speech | gpu_mldrift @ litert 2.2.0: 81.6% · 154 partition(s) | run_failed stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-09-28 | load_failed stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-09-28 | galaxy-s26 gpu_mldrift: output_mismatch · 4328.885 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (build BP4A.251205.006.S942QOPS1AZH9) · 2026-11-01galaxy-s26 gpu_mldrift: output_mismatch · 1426.072 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (build BP4A.251205.006.S942QOPS1AZH9) · 2026-11-02pixel-11-pro gpu_mldrift: output_mismatch · 3891.642 ms stale — latest known 2.3.0Pixel 11 Pro · Google Tensor G6 · litert 2.2.0 · Android 17 (build CD1A.260618.001.A7) · 2026-11-01pixel-11-pro gpu_mldrift: output_mismatch · 2166.519 ms stale — latest known 2.3.0Pixel 11 Pro · Google Tensor G6 · litert 2.2.0 · Android 17 (build CD1A.260905.001.B1) · 2026-11-02 | CARD.md · card.json | — |
audio8-tts-preview-0.6b__codec_decoder_int8_t128 | text-to-speech | gpu_mldrift @ litert 2.2.0: 90% · 74 partition(s) | run_failed stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-09-28 | load_failed stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-09-28 | galaxy-s26 cpu_xnnpack: fallback · 3978.71 msGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.2 · Android 16 (build BP4A.251205.006.S942QOPS1AZH9) · 2026-11-02galaxy-s26 cpu_xnnpack: fallback · 3112.992 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (build BP4A.251205.006.S942QOPS1AZH9) · 2026-11-02pixel-11-pro cpu_xnnpack: fallback · 5731.868 msPixel 11 Pro · Google Tensor G6 · litert 2.3.2 · Android 17 (build CD1A.260618.001.A7) · 2026-11-02 | CARD.md · card.json | — |
audio8-tts-preview-0.6b__codec_encoder_fp16_10s | text-to-speech | gpu_mldrift @ litert 2.2.0: 83.5% · 210 partition(s) | run_failed stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-09-28 | load_failed stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-09-28 | galaxy-s26 cpu_xnnpack: fallback · 3099.196 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (build BP4A.251205.006.S942QOPS1AZH9) · 2026-11-02galaxy-s26 gpu_mldrift: output_mismatch · 452.83 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (build BP4A.251205.006.S942QOPS1AZH9) · 2026-11-02pixel-11-pro cpu_xnnpack: fallback · 5240.949 ms stale — latest known 2.3.0Pixel 11 Pro · Google Tensor G6 · litert 2.2.0 · Android 17 (build CD1A.260618.001.A7) · 2026-11-01pixel-11-pro gpu_mldrift: output_mismatch · 1051.864 ms stale — latest known 2.3.0Pixel 11 Pro · Google Tensor G6 · litert 2.2.0 · Android 17 (build CD1A.260618.001.A7) · 2026-11-01 | CARD.md · card.json | — |
audio8-tts-preview-0.6b__fast_ar_int8 | text-to-speech | gpu_mldrift @ litert 2.2.0: 91.2% · 10 partition(s) | pass · 2.48 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-09-28 | output_mismatch · 3.423 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-09-28 | not measured | CARD.md · card.json | run it |
audio8-tts-preview-0.6b__slow_ar_int4 | text-to-speech | gpu_mldrift @ litert 2.2.0: 93.5% · 104 partition(s) | not measured | not measured | mac-studio-m4-max cpu_xnnpack: pass stale — latest known 2.3.0Mac Studio (M4 Max) · Apple M4 Max · litert 2.2.0 · macOS 27.0 (26A428) · 2026-09-28 | CARD.md · card.json | — |
audio8-tts-preview-0.6b__slow_ar_int8 | text-to-speech | gpu_mldrift @ litert 2.2.0: 93.5% · 104 partition(s) | not measured | not measured | galaxy-s26 cpu_xnnpack: pass stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (build BP4A.251205.006.S942QOPS1AZH9) · 2026-10-10mac-studio-m4-max cpu_xnnpack: pass stale — latest known 2.3.0Mac Studio (M4 Max) · Apple M4 Max · litert 2.2.0 · macOS 27.0 (26A428) · 2026-09-28mac-studio-m4-max cpu_xnnpack: pass stale — latest known 2.3.0Mac Studio (M4 Max) · Apple M4 Max · litert 2.2.0 · macOS 27.0 (26A428) · 2026-09-28 | CARD.md · card.json | — |
basic-pitch | audio-classification | — | pass · 24.417 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | pass · 6.148 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | galaxy-s26 gpu_mldrift: pass · 248.21 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 · 2026-08-25galaxy-s26 npu_qnn: pass · 4544.38 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · QAIRT (Hexagon, JIT on-device) · Android 16 · 2026-08-25raspberry-pi-5 cpu_xnnpack: pass · 23.973 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
bisenet-face-parsing | image-segmentation | — | pass · 442.142 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | pass · 79.592 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | raspberry-pi-5 cpu_xnnpack: pass · 249.388 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
bitcpm-cann-1b-int4 | text-generation | — | not measured | not measured | mac-studio-m4-max gpu: pass · 188.85 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.16.0 · 2026-08-12pixel-8a gpu: pass · 7.84 tok/s decodePixel 8a · Google Tensor G3 · litert-lm gallery-1.0.15 · 2026-07-23 | CARD.md · card.json | — |
clipseg-rd64__clipseg_decoder | image-segmentation | — | pass · 21.015 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | pass · 4.105 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | galaxy-s26 gpu_mldrift: pass · 3.252 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 · 2026-08-25galaxy-s26 npu_qnn: pass · 1.67 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · QAIRT (Hexagon, JIT on-device) · Android 16 · 2026-08-25raspberry-pi-5 cpu_xnnpack: pass · 19.04 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
clipseg-rd64__clipseg_text_fp16 | image-segmentation | — | pass · 370.805 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | pass · 8.715 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | galaxy-s26 gpu_mldrift: pass · 6.199 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 · 2026-08-26galaxy-s26 npu_qnn: pass · 2.1 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · QAIRT (Hexagon, JIT on-device) · Android 16 · 2026-08-26raspberry-pi-5 cpu_xnnpack: pass · 69.883 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
clipseg-rd64__clipseg_vision_fp16 | image-segmentation | — | pass · 4231.068 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | output_mismatch · 41.017 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | galaxy-s26 gpu_mldrift: pass · 35.509 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 · 2026-08-26galaxy-s26 npu_qnn: pass · 15.903 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · QAIRT (Hexagon, JIT on-device) · Android 16 · 2026-08-26raspberry-pi-5 cpu_xnnpack: pass · 856.718 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
cloth-segmentation-u2net | image-segmentation | — | pass · 9313.855 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | output_mismatch · 114.715 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | raspberry-pi-5 cpu_xnnpack: pass · 6544.635 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
cmgan | audio-to-audio | — | pass · 9213.145 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | output_mismatch · 114.805 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | galaxy-s26 gpu_mldrift: pass · 376.162 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 · 2026-08-26galaxy-s26 npu_qnn: load_failed stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · QAIRT (Hexagon, JIT on-device) · Android 16 · 2026-08-25raspberry-pi-5 cpu_xnnpack: pass · 7768.922 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
confucius4-r2t2 | automatic-speech-recognition | — | not measured | not measured | galaxy-s26 cpu: fallback stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.1 · Android 16 · 2026-10-02galaxy-s26 cpu: pass · 3480 ms stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.17.1 · Android 16 · 2026-10-10galaxy-s26 gpu: pass stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.1 · Android 16 · 2026-10-02mac-studio-m4-max cpu: pass · 1420 ms stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.17.1 · macOS 27.0 · 2026-10-02mac-studio-m4-max gpu: pass · 610 ms stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.17.1 · macOS 27.0 · 2026-10-02 | CARD.md · card.json | — |
cpga-net-lowlight | image-to-image | — | pass · 248.613 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | pass · 17.345 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | galaxy-s26 gpu_mldrift: pass · 14.94 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 · 2026-08-25galaxy-s26 npu_qnn: pass · 4.62 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · QAIRT (Hexagon, JIT on-device) · Android 16 · 2026-08-25raspberry-pi-5 cpu_xnnpack: pass · 78.796 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
crepe-pitch | audio-classification | — | pass · 188.315 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | pass · 6.043 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | raspberry-pi-5 cpu_xnnpack: pass · 49.534 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
d-fine-s__dfine_grapha_fp16 | object-detection | — | pass · 1461.235 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | pass · 47.127 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | galaxy-s26 gpu_mldrift: pass · 19.73 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 · 2026-08-25galaxy-s26 npu_qnn: pass · 23.812 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · QAIRT (Hexagon, JIT on-device) · Android 16 · 2026-08-25raspberry-pi-5 cpu_xnnpack: pass · 358.316 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
d-fine-s__dfine_graphb_fp16 | object-detection | — | pass · 432.95 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | output_mismatch · 25.665 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | galaxy-s26 gpu_mldrift: pass · 134.036 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 · 2026-08-25galaxy-s26 npu_qnn: pass · 172.363 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · QAIRT (Hexagon, JIT on-device) · Android 16 · 2026-08-25raspberry-pi-5 cpu_xnnpack: pass · 1018.76 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
decider-0.8b-fp16 | text-classification | — | not measured | not measured | galaxy-s26 cpu: fallback stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-23galaxy-s26 gpu: load_failed stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-23galaxy-s26 gpu: load_failed stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-23mac-studio-m4-max cpu: pass stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.17.1 · macOS 27.0 (26A428) · 2026-09-23 | CARD.md · card.json | — |
decider-0.8b-int8 | text-classification | — | not measured | not measured | galaxy-s26 cpu: fallback stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-23galaxy-s26 gpu: pass stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-23galaxy-s26 gpu: pass stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-23mac-studio-m4-max cpu: pass stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.17.1 · macOS 27.0 (26A428) · 2026-09-23 | CARD.md · card.json | — |
decider-2b-vision-fp16 | image-text-to-text | — | not measured | not measured | mac-studio-m4-max cpu: pass stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.17.1 · macOS 27.0 (26A428) · 2026-09-29mac-studio-m4-max cpu_xnnpack: pass stale — latest known 2.3.0Mac Studio (M4 Max) · Apple M4 Max · litert 2.2.0 · macOS 27.0 (26A428) · 2026-09-29mac-studio-m4-max gpu: pass stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.17.1 · macOS 27.0 (26A428) · 2026-09-29 | CARD.md · card.json | — |
decider-2b-vision-fp16-int8vocab | image-text-to-text | — | not measured | not measured | mac-studio-m4-max cpu: pass stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.17.1 · macOS 27.0 (26A428) · 2026-09-29mac-studio-m4-max cpu_xnnpack: pass stale — latest known 2.3.0Mac Studio (M4 Max) · Apple M4 Max · litert 2.2.0 · macOS 27.0 (26A428) · 2026-09-29mac-studio-m4-max gpu: pass stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.17.1 · macOS 27.0 (26A428) · 2026-09-29 | CARD.md · card.json | — |
decider-2b-vision-int8 | image-text-to-text | — | not measured | not measured | galaxy-s26 cpu: fallback stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.1 · Android 16 · 2026-09-29galaxy-s26 cpu: pass stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.1 · Android 16 (build BP4A.251205.006.S942QOPS1AZH9) · 2026-10-10galaxy-s26 cpu: fallback stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-29galaxy-s26 gpu: pass stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.1 · Android 16 · 2026-09-29galaxy-s26 gpu: pass stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.1 · Android 16 (build BP4A.251205.006.S942QOPS1AZH9) · 2026-10-10galaxy-s26 gpu: pass stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-29mac-studio-m4-max cpu: pass stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.17.1 · macOS 27.0 (26A428) · 2026-09-29mac-studio-m4-max gpu: pass stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.17.1 · macOS 27.0 (26A428) · 2026-09-29 | CARD.md · card.json | — |
deepseek-r1-distill-qwen-7b-q4-block32-ekv4096 | text-generation | — | not measured | not measured | galaxy-s26 cpu: fallback · 8.81 tok/s decodeGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm prebuilt-adac974b · Android 16 · 2026-10-05galaxy-s26 cpu: fallback · 8.38 tok/s decodeGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm prebuilt-adac974b · Android 16 · 2026-10-05galaxy-s26 gpu: pass · 12.26 tok/s decodeGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm prebuilt-adac974b · Android 16 · 2026-10-05galaxy-s26 gpu: pass · 8.43 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-08-24galaxy-s26 gpu: pass · 8.57 tok/s decodeGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm prebuilt-adac974b · Android 16 · 2026-10-05mac-studio-m4-max cpu: fallback · 12.25 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.17.1 · 2026-10-05mac-studio-m4-max cpu: fallback · 18.24 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.17.1 · 2026-10-05mac-studio-m4-max gpu: pass · 76.91 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.17.1 · 2026-10-05mac-studio-m4-max gpu: pass · 79.12 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.17.1 · 2026-10-05raspberry-pi-5 cpu: pass · 1.76 tok/s decode stale — latest known 0.19.1Raspberry Pi 5 Model B Rev 1.1 · Broadcom BCM2712 · litert-lm 0.16.1 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-09-01 | CARD.md · card.json | — |
dehazeformer-mct | image-to-image | — | pass · 709.245 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | pass · 96.34 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | raspberry-pi-5 cpu_xnnpack: pass · 2183.792 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
dewarpnet | image-to-image | — | pass · 633.472 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | pass · 22.56 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | raspberry-pi-5 cpu_xnnpack: pass · 593.957 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
dinov2-vit-s14 | image-feature-extraction | — | pass · 2977.405 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | output_mismatch · 29.86 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | not measured | CARD.md · card.json | run it |
dis-isnet | image-segmentation | — | pass · 4440.095 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | pass · 65.37 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | raspberry-pi-5 cpu_xnnpack: pass · 2956.654 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
dm-count-crowd | image-to-image | — | pass · 2999.897 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | pass · 14.847 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | raspberry-pi-5 cpu_xnnpack: pass · 1923.847 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
edsr-x4 | image-to-image | — | pass · 1531.135 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | pass · 31.675 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | raspberry-pi-5 cpu_xnnpack: pass · 1015.911 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
ema-lightning__ema_align_multi_fp32 | text-to-speech | gpu_mldrift @ litert 2.2.0: 92.5% · 15 partition(s) | not measured | not measured | galaxy-s26 cpu_xnnpack: pass · 0.2490625 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (build BP4A.251205.006.S942QOPS1AZH9) · 2026-10-11galaxy-s26 cpu_xnnpack: pass · 0.22031250000000002 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (build BP4A.251205.006.S942QOPS1AZH9) · 2026-10-11galaxy-s26 gpu_mldrift: output_mismatch · 0.6097915 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (build BP4A.251205.006.S942QOPS1AZH9) · 2026-10-11galaxy-s26 gpu_mldrift: pass · 0.6510415 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (build BP4A.251205.006.S942QOPS1AZH9) · 2026-10-11galaxy-s26 npu_qnn: output_mismatch · 0.487 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Qualcomm QNN HTP runtime libraries staged beside the runner (libQnnHtp.so, libQnnHtpPrepare.so, libQnnHtpV81Skel.so / Stub.so, libQnnSystem.so; sha256 MATCH) with LiteRT's libLiteRtCompilerPlugin_Qualcomm.so / libLiteRtDispatch_Qualcomm.so; no version string is printed by the runner · Android 16 (build BP4A.251205.006.S942QOPS1AZH9) · 2026-10-11mac-studio-m4-max cpu_xnnpack: pass · 0.55 ms stale — latest known 2.3.0Mac Studio (M4 Max) · Apple M4 Max · litert 2.2.0 · macOS 27.0 (26A428) · 2026-10-11mac-studio-m4-max cpu_xnnpack: pass · 1.01 ms stale — latest known 2.3.0Mac Studio (M4 Max) · Apple M4 Max · litert 2.2.0 · macOS 27.0 (26A428) · 2026-10-11mac-studio-m4-max cpu_xnnpack: pass · 1.17 ms stale — latest known 2.3.0Mac Studio (M4 Max) · Apple M4 Max · litert 2.2.0 · macOS 27.0 (26A428) · 2026-10-11mac-studio-m4-max cpu_xnnpack: pass · 0.18 ms stale — latest known 2.3.0Mac Studio (M4 Max) · Apple M4 Max · litert 2.2.0 · macOS 27.0 (26A428) · 2026-10-11mac-studio-m4-max cpu_xnnpack: pass · 0.33 ms stale — latest known 2.3.0Mac Studio (M4 Max) · Apple M4 Max · litert 2.2.0 · macOS 27.0 (26A428) · 2026-10-11mac-studio-m4-max cpu_xnnpack: pass stale — latest known 2.3.0Mac Studio (M4 Max) · Apple M4 Max · litert 2.2.0 · macOS 27.0 (26A428) · 2026-10-11 | CARD.md · card.json | — |
ema-lightning__ema_decoder_multi_fp16 | text-to-speech | gpu_mldrift @ litert 2.2.0: 76.7% · 442 partition(s) | not measured | not measured | galaxy-s26 cpu_xnnpack: pass · 99.552291 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (build BP4A.251205.006.S942QOPS1AZH9) · 2026-10-11galaxy-s26 gpu_mldrift: output_mismatch · 30.506614 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (build BP4A.251205.006.S942QOPS1AZH9) · 2026-10-11galaxy-s26 gpu_mldrift: output_mismatch · 14.624141000000002 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (build BP4A.251205.006.S942QOPS1AZH9) · 2026-10-11mac-studio-m4-max cpu_xnnpack: pass stale — latest known 2.3.0Mac Studio (M4 Max) · Apple M4 Max · litert 2.2.0 · macOS 27.0 (26A428) · 2026-10-11 | CARD.md · card.json | — |
ema-lightning__ema_step_multi_fp32 | text-to-speech | gpu_mldrift @ litert 2.2.0: 96.4% · 100 partition(s) | not measured | not measured | galaxy-s26 cpu_xnnpack: pass · 14.2302605 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (build BP4A.251205.006.S942QOPS1AZH9) · 2026-10-11galaxy-s26 cpu_xnnpack: pass · 2.472969 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (build BP4A.251205.006.S942QOPS1AZH9) · 2026-10-11galaxy-s26 cpu_xnnpack: pass · 2.5629429999999997 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (build BP4A.251205.006.S942QOPS1AZH9) · 2026-10-11galaxy-s26 gpu_mldrift: output_mismatch · 2.4051825 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (build BP4A.251205.006.S942QOPS1AZH9) · 2026-10-11galaxy-s26 gpu_mldrift: pass · 3.6925 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (build BP4A.251205.006.S942QOPS1AZH9) · 2026-10-11galaxy-s26 gpu_mldrift: pass · 3.2139325000000003 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (build BP4A.251205.006.S942QOPS1AZH9) · 2026-10-11galaxy-s26 npu_qnn: output_mismatch · 1.557 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Qualcomm QNN HTP runtime libraries staged beside the runner (libQnnHtp.so, libQnnHtpPrepare.so, libQnnHtpV81Skel.so / Stub.so, libQnnSystem.so; sha256 MATCH) with LiteRT's libLiteRtCompilerPlugin_Qualcomm.so / libLiteRtDispatch_Qualcomm.so; no version string is printed by the runner · Android 16 (build BP4A.251205.006.S942QOPS1AZH9) · 2026-10-11mac-studio-m4-max cpu_xnnpack: pass · 3.97 ms stale — latest known 2.3.0Mac Studio (M4 Max) · Apple M4 Max · litert 2.2.0 · macOS 27.0 (26A428) · 2026-10-11mac-studio-m4-max cpu_xnnpack: pass · 6.98 ms stale — latest known 2.3.0Mac Studio (M4 Max) · Apple M4 Max · litert 2.2.0 · macOS 27.0 (26A428) · 2026-10-11mac-studio-m4-max cpu_xnnpack: pass · 1.97 ms stale — latest known 2.3.0Mac Studio (M4 Max) · Apple M4 Max · litert 2.2.0 · macOS 27.0 (26A428) · 2026-10-11mac-studio-m4-max cpu_xnnpack: pass · 9.57 ms stale — latest known 2.3.0Mac Studio (M4 Max) · Apple M4 Max · litert 2.2.0 · macOS 27.0 (26A428) · 2026-10-11mac-studio-m4-max cpu_xnnpack: pass · 2.76 ms stale — latest known 2.3.0Mac Studio (M4 Max) · Apple M4 Max · litert 2.2.0 · macOS 27.0 (26A428) · 2026-10-11mac-studio-m4-max cpu_xnnpack: pass stale — latest known 2.3.0Mac Studio (M4 Max) · Apple M4 Max · litert 2.2.0 · macOS 27.0 (26A428) · 2026-10-11 | CARD.md · card.json | — |
ema-lightning__ema_text_multi_fp32 | text-to-speech | gpu_mldrift @ litert 2.2.0: 89.9% · 60 partition(s) | not measured | not measured | galaxy-s26 cpu_xnnpack: fallback · 0.655286 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (build BP4A.251205.006.S942QOPS1AZH9) · 2026-10-11galaxy-s26 cpu_xnnpack: fallback · 0.566771 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (build BP4A.251205.006.S942QOPS1AZH9) · 2026-10-11galaxy-s26 gpu_mldrift: output_mismatch stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (build BP4A.251205.006.S942QOPS1AZH9) · 2026-10-11galaxy-s26 gpu_mldrift: pass · 1.3771875 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (build BP4A.251205.006.S942QOPS1AZH9) · 2026-10-11galaxy-s26 gpu_mldrift: pass · 1.850495 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (build BP4A.251205.006.S942QOPS1AZH9) · 2026-10-11galaxy-s26 npu_qnn: output_mismatch · 0.7004999999999999 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Qualcomm QNN HTP runtime libraries staged beside the runner (libQnnHtp.so, libQnnHtpPrepare.so, libQnnHtpV81Skel.so / Stub.so, libQnnSystem.so; sha256 MATCH) with LiteRT's libLiteRtCompilerPlugin_Qualcomm.so / libLiteRtDispatch_Qualcomm.so; no version string is printed by the runner · Android 16 (build BP4A.251205.006.S942QOPS1AZH9) · 2026-10-11mac-studio-m4-max cpu_xnnpack: pass · 0.84 ms stale — latest known 2.3.0Mac Studio (M4 Max) · Apple M4 Max · litert 2.2.0 · macOS 27.0 (26A428) · 2026-10-11mac-studio-m4-max cpu_xnnpack: pass · 1.27 ms stale — latest known 2.3.0Mac Studio (M4 Max) · Apple M4 Max · litert 2.2.0 · macOS 27.0 (26A428) · 2026-10-11mac-studio-m4-max cpu_xnnpack: pass · 0.38 ms stale — latest known 2.3.0Mac Studio (M4 Max) · Apple M4 Max · litert 2.2.0 · macOS 27.0 (26A428) · 2026-10-11mac-studio-m4-max cpu_xnnpack: pass · 0.56 ms stale — latest known 2.3.0Mac Studio (M4 Max) · Apple M4 Max · litert 2.2.0 · macOS 27.0 (26A428) · 2026-10-11mac-studio-m4-max cpu_xnnpack: pass stale — latest known 2.3.0Mac Studio (M4 Max) · Apple M4 Max · litert 2.2.0 · macOS 27.0 (26A428) · 2026-10-11 | CARD.md · card.json | — |
example-tiny-clean | image-classification | gpu_mldrift @ litert 0.0.0-example: 100% · 1 partition(s) | not measured | not measured | example-phone cpu_xnnpack: pass · 4.8 ms exampleExample Phone · Example SoC · litert 0.0.0-example · EXAMPLE.240001.001 · 2026-01-01example-phone npu_qnn: fallback · 1.2 ms exampleExample Phone · Example SoC · litert 0.0.0-example · QNN SDK 0.0.0-example · EXAMPLE.240001.001 · 2026-01-01 | CARD.md · card.json | — |
example-tiny-mixed | image-segmentation | gpu_mldrift @ litert 0.0.0-example: 40% · 1 partition(s) | not measured | not measured | example-workstation gpu: pass · 45.6 tok/s decode exampleExample Workstation · Example Desktop SoC · litert-lm 0.0.0-example · macOS 0.0-example · 2026-01-02 | CARD.md · card.json | — |
example-web-a | example-elementwise | — | pass · 0.032 ms example stale — latest known 2.6.0example-dev-mac-arm64 · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-10 | pass · 0.55 ms example stale — latest known 2.6.0example-dev-mac-arm64 · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-10 | not measured | CARD.md · card.json | run it |
example-web-b | example-elementwise | — | pass · 0.043 ms example stale — latest known 2.6.0example-dev-mac-arm64 · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-10 | pass · 0.635 ms example stale — latest known 2.6.0example-dev-mac-arm64 · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-10 | not measured | CARD.md · card.json | run it |
falcon-h1-0.5b-instruct-int8 | text-generation | — | not measured | not measured | galaxy-s26 cpu: fallback · 40.48 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-05galaxy-s26 gpu: pass · 28.14 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-08-24pixel-8a cpu: fallback · 25.86 tok/s decode stale — latest known 0.19.1Pixel 8a · Tensor G3 · litert-lm 0.16.0 · Android 16 · 2026-09-05raspberry-pi-5 cpu: pass · 7 tok/s decode stale — latest known 0.19.1Raspberry Pi 5 Model B Rev 1.1 · Broadcom BCM2712 · litert-lm 0.16.1 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-09-02 | CARD.md · card.json | — |
falcon-h1-1.5b-deep-instruct-int8 | text-generation | — | not measured | not measured | galaxy-s26 cpu: fallback · 11.97 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-05galaxy-s26 gpu: pass · 6 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-08-24raspberry-pi-5 cpu: pass · 2.05 tok/s decode stale — latest known 0.19.1Raspberry Pi 5 Model B Rev 1.1 · Broadcom BCM2712 · litert-lm 0.16.1 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-09-02 | CARD.md · card.json | — |
falcon-h1-1.5b-instruct-int8 | text-generation | — | not measured | not measured | galaxy-s26 cpu: fallback · 20.71 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-05galaxy-s26 gpu: pass · 20.6 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-08-24raspberry-pi-5 cpu: pass · 3.38 tok/s decode stale — latest known 0.19.1Raspberry Pi 5 Model B Rev 1.1 · Broadcom BCM2712 · litert-lm 0.16.1 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-09-02 | CARD.md · card.json | — |
falcon-h1-tiny-r-0.6b-int8 | text-generation | — | not measured | not measured | galaxy-s26 cpu: fallback · 23.9 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-05galaxy-s26 gpu: pass · 16.13 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-05iphone-17-pro cpu: pass · 29.38 tok/s decode stale — latest known 0.19.1iPhone 17 Pro · A19-Pro · litert-lm 0.16.0 · iOS 27.0 · 2026-09-01iphone-17-pro gpu: pass · 22.84 tok/s decode stale — latest known 0.19.1iPhone 17 Pro · A19-Pro · litert-lm 0.16.0 · iOS 27.0 · 2026-09-01mac-studio-m4-max cpu: pass · 47.75 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · litert-lm 0.16.0 · 2026-09-01mac-studio-m4-max gpu: pass · 97.76 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · litert-lm 0.16.0 · 2026-09-01pixel-8a cpu: fallback · 14.02 tok/s decode stale — latest known 0.19.1Pixel 8a · Tensor G3 · litert-lm 0.16.0 · Android 16 · 2026-09-05pixel-8a gpu: pass · 9.98 tok/s decode stale — latest known 0.19.1Pixel 8a · Tensor G3 · litert-lm 0.16.0 · Android 16 · 2026-09-05 | CARD.md · card.json | — |
falcon-h1r-7b-int4 | text-generation | — | not measured | not measured | mac-studio-m4-max cpu: pass · 12.35 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.18.0 · macOS 27.0 (26A428) · 2026-10-15mac-studio-m4-max gpu: pass · 39.705 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.18.0 · macOS 27.0 (26A428) · 2026-10-15pixel-8a gpu: load_failedPixel 8a · Google Tensor G3 · litert-lm prebuilt-adac974b · Android 16 (CP1A.260505.005) · 2026-10-15 | CARD.md · card.json | — |
falcon-h1r-7b-int8 | text-generation | — | not measured | not measured | mac-studio-m4-max cpu: pass · 11.585 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.18.0 · macOS 27.0 (26A428) · 2026-10-15mac-studio-m4-max gpu: pass stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.17.1 · macOS 27.0 (26A428) · 2026-10-14mac-studio-m4-max gpu: pass · 35.875 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.18.0 · macOS 27.0 (26A428) · 2026-10-15 | CARD.md · card.json | — |
falconh1-3b | text-generation | — | not measured | not measured | galaxy-s26 gpu: pass · 11.49 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-08-24iphone-17-pro cpu: pass · 7.85 tok/s decode stale — latest known 0.19.1iPhone 17 Pro · litert-lm 0.16.0 · iOS 27.0 · 2026-08-14iphone-17-pro gpu: pass · 14.04 tok/s decode stale — latest known 0.19.1iPhone 17 Pro · litert-lm 0.16.0 · iOS 27.0 · 2026-08-14mac-m4-max cpu: pass · 20.88 tok/s decode stale — latest known 0.19.1mac-m4-max · litert-lm 0.16.0 · 2026-08-14mac-m4-max gpu: pass · 65.32 tok/s decode stale — latest known 0.19.1mac-m4-max · litert-lm 0.16.0 · 2026-08-14raspberry-pi-5 cpu: pass · 1.9 tok/s decode stale — latest known 0.19.1Raspberry Pi 5 Model B Rev 1.1 · Broadcom BCM2712 · litert-lm 0.16.1 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-09-02 | CARD.md · card.json | — |
fast-neural-style__style_candy_fp16 | image-to-image | — | pass · 1307.595 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | output_mismatch · 7.34 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | galaxy-s26 gpu_mldrift: pass · 13.875 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 · 2026-08-25galaxy-s26 npu_qnn: pass · 7.405 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · QAIRT (Hexagon, JIT on-device) · Android 16 · 2026-08-25raspberry-pi-5 cpu_xnnpack: pass · 287.492 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
fast-neural-style__style_mosaic_fp16 | image-to-image | — | pass · 1307.638 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | output_mismatch · 7.19 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | galaxy-s26 gpu_mldrift: pass · 13.845 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 · 2026-08-25galaxy-s26 npu_qnn: pass · 7.414 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · QAIRT (Hexagon, JIT on-device) · Android 16 · 2026-08-25raspberry-pi-5 cpu_xnnpack: pass · 288.032 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
fast-neural-style__style_rain_princess_fp16 | image-to-image | — | pass · 1307.967 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | output_mismatch · 7.342 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | galaxy-s26 gpu_mldrift: pass · 13.845 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 · 2026-08-25galaxy-s26 npu_qnn: pass · 7.481 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · QAIRT (Hexagon, JIT on-device) · Android 16 · 2026-08-25raspberry-pi-5 cpu_xnnpack: pass · 287.576 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
fast-neural-style__style_udnie_fp16 | image-to-image | — | pass · 1307.873 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | output_mismatch · 7.373 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | galaxy-s26 gpu_mldrift: pass · 13.764 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 · 2026-08-25galaxy-s26 npu_qnn: pass · 7.415 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · QAIRT (Hexagon, JIT on-device) · Android 16 · 2026-08-25raspberry-pi-5 cpu_xnnpack: pass · 287.501 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
fastcontext-1.0-4b-sft-model-block128 | text-generation | — | not measured | not measured | galaxy-s26 gpu: pass · 7.49 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-08-24raspberry-pi-5 cpu: pass · 1.49 tok/s decode stale — latest known 0.19.1Raspberry Pi 5 Model B Rev 1.1 · Broadcom BCM2712 · litert-lm 0.16.1 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-09-02 | CARD.md · card.json | — |
fastcontext-4b | text-generation | — | not measured | not measured | galaxy-s26 cpu: fallback · 7.14 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-05mac-studio-m4-max gpu: pass · 74.53 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.14.0 · 2026-07-23raspberry-pi-5 cpu: pass · 1.44 tok/s decode stale — latest known 0.19.1Raspberry Pi 5 Model B Rev 1.1 · Broadcom BCM2712 · litert-lm 0.16.1 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-09-02 | CARD.md · card.json | — |
fun-asr-nano-2512 | automatic-speech-recognition | — | not measured | not measured | galaxy-s26 cpu: pass · 1237.1 ms stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.17.1 · Android 16 · 2026-09-27galaxy-s26 cpu: pass · 1830 ms stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.17.1 · Android 16 · 2026-10-10galaxy-s26 cpu: fallback · 31.55 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.1 · Android 16 · 2026-09-27galaxy-s26 gpu: pass · 1690 ms stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.17.1 · Android 16 (build BP4A.251205.006.S942QOPS1AZH9) · 2026-10-09galaxy-s26 gpu: pass · 26.26 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.1 · Android 16 · 2026-09-27mac-studio-m4-max cpu: pass · 48.94 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.17.1 · macOS 27.0 · 2026-09-27mac-studio-m4-max cpu: pass · 50.21 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.17.1 · macOS 27.0 · 2026-09-27mac-studio-m4-max gpu: pass · 143.03 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.17.1 · macOS 27.0 · 2026-09-27mac-studio-m4-max gpu: pass · 137.18 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.17.1 · macOS 27.0 · 2026-09-27 | CARD.md · card.json | — |
g9v3-3b-int4 | text-generation | — | not measured | not measured | galaxy-s26 gpu: pass · 18.465 tok/s decodeGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm prebuilt-adac974b · Android 16 (BP4A.251205.006) · 2026-10-13mac-studio-m4-max cpu: pass stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.17.1 · macOS 27.0 · 2026-10-13mac-studio-m4-max gpu: pass · 61.485 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.18.0 · macOS 27.0 · 2026-10-13pixel-8a cpu: fallback · 5.785 tok/s decodePixel 8a · Tensor G3 · litert-lm prebuilt-adac974b · Android 16 (CP1A.260505.005) · 2026-10-18pixel-8a gpu: pass · 8.955 tok/s decodePixel 8a · Tensor G3 · litert-lm prebuilt-adac974b · Android 16 (CP1A.260505.005) · 2026-10-13 | CARD.md · card.json | — |
g9v3-3b-int8 | text-generation | — | not measured | not measured | galaxy-s26 cpu: fallback · 9.925 tok/s decodeGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm prebuilt-adac974b · Android 16 (BP4A.251205.006) · 2026-10-13galaxy-s26 gpu: pass · 12.225 tok/s decodeGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm prebuilt-adac974b · Android 16 (BP4A.251205.006) · 2026-10-13mac-studio-m4-max cpu: pass stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.17.1 · macOS 27.0 · 2026-10-18mac-studio-m4-max cpu: pass · 24.86 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.18.0 · macOS 27.0 · 2026-10-17mac-studio-m4-max gpu: pass · 84.17 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.18.0 · macOS 27.0 · 2026-10-17pixel-8a cpu: fallback · 4.68 tok/s decodePixel 8a · Tensor G3 · litert-lm prebuilt-adac974b · Android 16 (CP1A.260505.005) · 2026-10-18pixel-8a gpu: load_failedPixel 8a · Tensor G3 · litert-lm prebuilt-adac974b · Android 16 (CP1A.260505.005) · 2026-10-13 | CARD.md · card.json | — |
gfpgan-v1.4__gfpgan_fp16 | image-to-image | — | pass · 10.412 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | output_mismatch · 60.61 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | galaxy-s26 gpu_mldrift: pass · 88.949 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 · 2026-08-26galaxy-s26 npu_qnn: pass · 32.258 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · QAIRT (Hexagon, JIT on-device) · Android 16 · 2026-08-26raspberry-pi-5 cpu_xnnpack: pass · 4085.305 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
gfpgan-v1.4__yunet_fp16 | image-to-image | — | not measured | not measured | galaxy-s26 gpu_mldrift: pass · 3.277 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 · 2026-08-25galaxy-s26 npu_qnn: pass · 1.828 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · QAIRT (Hexagon, JIT on-device) · Android 16 · 2026-08-25raspberry-pi-5 cpu_xnnpack: pass · 33.537 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | — |
gliclass-edge-v3.0__gliclass_edge_v3_s128_fp32 | zero-shot-classification | gpu_mldrift @ litert 2.2.0: 96.9% · 22 partition(s) | pass · 48.098 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-10-02 | pass · 5.635 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-10-02 | galaxy-s26 cpu_xnnpack: pass · 11.875052 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (build BP4A.251205.006.S942QOPS1AZH9) · 2026-10-03galaxy-s26 cpu_xnnpack: pass · 7.8285420000000006 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (SDK 36) · 2026-10-02galaxy-s26 cpu_xnnpack: pass · 6.767916 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (SDK 36) · 2026-10-02galaxy-s26 gpu_mldrift: output_mismatch · 3.864323 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (SDK 36) · 2026-10-02galaxy-s26 gpu_mldrift: pass · 5.860834 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (SDK 36) · 2026-10-02galaxy-s26 gpu_mldrift: output_mismatch · 5.805 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (SDK 36) · 2026-10-02galaxy-s26 gpu_mldrift: pass · 6.165365 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (build BP4A.251205.006.S942QOPS1AZH9) · 2026-10-04 | CARD.md · card.json | run it |
gliclass-edge-v3.0__gliclass_edge_v3_s128_wfp16 | zero-shot-classification | gpu_mldrift @ litert 2.2.0: 86.8% · 100 partition(s) | pass · 48.648 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-10-02 | pass · 5.575 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-10-02 | galaxy-s26 gpu_mldrift: output_mismatch · 3.8475 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (SDK 36) · 2026-10-02galaxy-s26 gpu_mldrift: output_mismatch · 5.844427 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (SDK 36) · 2026-10-02galaxy-s26 npu_qnn: output_mismatch · 2.51302 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (SDK 36) · 2026-10-02 | CARD.md · card.json | run it |
gliclass-edge-v3.0__gliclass_edge_v3_s256_fp32 | zero-shot-classification | gpu_mldrift @ litert 2.2.0: 96.9% · 22 partition(s) | pass · 101.72 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-10-02 | pass · 7.918 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-10-02 | galaxy-s26 cpu_xnnpack: pass · 24.105937 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (build BP4A.251205.006.S942QOPS1AZH9) · 2026-10-04galaxy-s26 cpu_xnnpack: pass · 15.649375 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (SDK 36) · 2026-10-02galaxy-s26 gpu_mldrift: pass · 8.230103999999999 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (SDK 36) · 2026-10-02galaxy-s26 gpu_mldrift: pass · 8.322448 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (SDK 36) · 2026-10-02galaxy-s26 gpu_mldrift: pass · 10.000468 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (build BP4A.251205.006.S942QOPS1AZH9) · 2026-10-04 | CARD.md · card.json | run it |
gliclass-edge-v3.0__gliclass_edge_v3_s256_wfp16 | zero-shot-classification | gpu_mldrift @ litert 2.2.0: 86.8% · 100 partition(s) | pass · 108.702 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-10-02 | pass · 8.063 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-10-02 | galaxy-s26 gpu_mldrift: output_mismatch · 8.332084 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (SDK 36) · 2026-10-02 | CARD.md · card.json | run it |
gliformer-large-ner__gliformer_large_ner_s128_wfp16 | token-classification | gpu_mldrift @ litert 2.2.0: 83.6% · 681 partition(s) | run_failed stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-09-26 | load_failed stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-09-26 | galaxy-s26 cpu_xnnpack: fallback · 322.54960900000003 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (SDK 36, build BP4A.251205.006) · 2026-09-26galaxy-s26 cpu_xnnpack: fallback · 189.904922 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (SDK 36, build BP4A.251205.006) · 2026-09-26galaxy-s26 gpu_mldrift: output_mismatch · 46.997577500000006 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (SDK 36, build BP4A.251205.006) · 2026-09-26galaxy-s26 gpu_mldrift: pass · 82.27380199999999 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (SDK 36, build BP4A.251205.006) · 2026-09-26galaxy-s26 gpu_mldrift: output_mismatch · 116.21625 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (SDK 36, build BP4A.251205.006) · 2026-09-26 | CARD.md · card.json | — |
gliformer-large-ner__gliformer_large_ner_s256_encoder_wfp16 | token-classification | gpu_mldrift @ litert 2.2.0: 89.1% · 194 partition(s) | run_failed stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-09-26 | load_failed stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-09-26 | galaxy-s26 gpu_mldrift: output_mismatch · 175.55677 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (SDK 36, build BP4A.251205.006) · 2026-09-26galaxy-s26 gpu_mldrift: pass · 390.950938 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (SDK 36, build BP4A.251205.006) · 2026-09-26 | CARD.md · card.json | — |
gliformer-large-ner__gliformer_large_ner_s256_head_wfp16 | token-classification | gpu_mldrift @ litert 2.2.0: 79.6% · 2568 partition(s) | pass · 2425.31 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-09-26 | load_failed stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-09-26 | galaxy-s26 cpu_xnnpack: fallback · 167.031874 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (SDK 36, build BP4A.251205.006) · 2026-09-26galaxy-s26 cpu_xnnpack: fallback · 128.378021 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (SDK 36, build BP4A.251205.006) · 2026-09-26 | CARD.md · card.json | run it |
gliformer-large-ner__gliformer_large_ner_s512_encoder_wfp16 | token-classification | gpu_mldrift @ litert 2.2.0: 89.1% · 194 partition(s) | run_failed stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-09-26 | load_failed stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-09-26 | galaxy-s26 gpu_mldrift: output_mismatch · 931.4937235 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (SDK 36, build BP4A.251205.006) · 2026-09-26 | CARD.md · card.json | — |
gliformer-large-ner__gliformer_large_ner_s512_head_wfp16 | token-classification | gpu_mldrift @ litert 2.2.0: 79.6% · 5128 partition(s) | run_failed stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-09-26 | load_failed stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-09-26 | galaxy-s26 cpu_xnnpack: fallback · 348.8584375 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (SDK 36, build BP4A.251205.006) · 2026-09-26 | CARD.md · card.json | — |
gliner2.5-decide__gliner25_decide_s128_wfp16 | text-classification | gpu_mldrift @ litert 2.2.0: 89% · 196 partition(s) | run_failed stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-09-26 | load_failed stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-09-26 | galaxy-s26 cpu_xnnpack: pass · 213.804844 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (SDK 36, build BP4A.251205.006) · 2026-09-26galaxy-s26 cpu_xnnpack: pass · 239.478073 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (build BP4A.251205.006.S942QOPS1AZH9) · 2026-10-03galaxy-s26 gpu_mldrift: pass · 69.639375 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (SDK 36, build BP4A.251205.006) · 2026-09-26galaxy-s26 gpu_mldrift: pass · 83.143021 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (SDK 36, build BP4A.251205.006) · 2026-09-26galaxy-s26 gpu_mldrift: pass · 69.971354 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (build BP4A.251205.006.S942QOPS1AZH9) · 2026-10-04 | CARD.md · card.json | — |
gliner2.5-decide__gliner25_decide_s256_wfp16 | text-classification | gpu_mldrift @ litert 2.2.0: 89% · 196 partition(s) | run_failed stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-09-26 | load_failed stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-09-26 | galaxy-s26 cpu_xnnpack: pass · 509.982031 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (SDK 36, build BP4A.251205.006) · 2026-09-26galaxy-s26 cpu_xnnpack: pass · 474.6950255 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (SDK 36, build BP4A.251205.006) · 2026-09-26galaxy-s26 cpu_xnnpack: pass · 556.168854 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (build BP4A.251205.006.S942QOPS1AZH9) · 2026-10-03galaxy-s26 gpu_mldrift: pass · 198.833698 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (SDK 36, build BP4A.251205.006) · 2026-09-26galaxy-s26 gpu_mldrift: pass · 205.67757749999998 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (SDK 36, build BP4A.251205.006) · 2026-09-26galaxy-s26 gpu_mldrift: pass · 184.383958 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (build BP4A.251205.006.S942QOPS1AZH9) · 2026-10-03 | CARD.md · card.json | — |
gliner2.5-decide__gliner25_decide_s512_wfp16 | text-classification | gpu_mldrift @ litert 2.2.0: 89% · 196 partition(s) | run_failed stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-09-26 | load_failed stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-09-26 | galaxy-s26 cpu_xnnpack: pass · 1665.9826295 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (SDK 36, build BP4A.251205.006) · 2026-09-26galaxy-s26 cpu_xnnpack: pass · 1242.723984 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (SDK 36, build BP4A.251205.006) · 2026-09-26galaxy-s26 gpu_mldrift: pass · 733.1154160000001 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (SDK 36, build BP4A.251205.006) · 2026-09-26galaxy-s26 gpu_mldrift: pass · 639.23664 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (SDK 36, build BP4A.251205.006) · 2026-09-26galaxy-s26 gpu_mldrift: pass · 582.360208 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (build BP4A.251205.006.S942QOPS1AZH9) · 2026-10-04galaxy-s26 gpu_mldrift: pass · 578.269948 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (build BP4A.251205.006.S942QOPS1AZH9) · 2026-10-04 | CARD.md · card.json | — |
gliner2.5-multi__gliner25_multi_s128_wfp16 | token-classification | gpu_mldrift @ litert 2.2.0: 89% · 127 partition(s) | pass · 1308.485 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-10-02 | output_mismatch · 21.68 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-10-02 | galaxy-s26 cpu_xnnpack: output_mismatch · 49.4759895 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (SDK 36) · 2026-10-02galaxy-s26 gpu_mldrift: output_mismatch · 15.035599000000001 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (SDK 36) · 2026-10-02galaxy-s26 gpu_mldrift: output_mismatch · 24.286589 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (SDK 36) · 2026-10-02galaxy-s26 npu_qnn: output_mismatch · 8.253828 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (SDK 36) · 2026-10-02 | CARD.md · card.json | run it |
gliner2.5-multi__gliner25_multi_s256_wfp16 | token-classification | gpu_mldrift @ litert 2.2.0: 89% · 127 partition(s) | pass · 2648.487 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-10-02 | output_mismatch · 30.443 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-10-02 | galaxy-s26 cpu_xnnpack: output_mismatch · 128.523334 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (SDK 36) · 2026-10-02galaxy-s26 gpu_mldrift: output_mismatch · 92.95927 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (SDK 36) · 2026-10-02galaxy-s26 npu_qnn: output_mismatch · 34.443021 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (SDK 36) · 2026-10-02 | CARD.md · card.json | run it |
gliner2.5-multi__gliner25_multi_s512_wfp16 | token-classification | gpu_mldrift @ litert 2.2.0: 89% · 127 partition(s) | run_failed stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-10-02 | pass · 50.885 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-10-02 | galaxy-s26 cpu_xnnpack: output_mismatch · 407.261745 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (SDK 36) · 2026-10-02galaxy-s26 gpu_mldrift: output_mismatch · 422.52502549999997 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (SDK 36) · 2026-10-02galaxy-s26 npu_qnn: output_mismatch · 186.5391665 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (SDK 36) · 2026-10-02 | CARD.md · card.json | run it |
gpt-oss-120b-mixed-int4-ekv4096 | text-generation | — | not measured | not measured | mac-studio-m4-max cpu: passMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.1 · macOS 27.0 (26A428) · 2026-11-10mac-studio-m4-max cpu: passMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.1 · macOS 27.0 (26A428) · 2026-11-10mac-studio-m4-max cpu: pass · 4.75 tok/s decodeMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.1 · macOS 27.0 (26A428) · 2026-11-10 | CARD.md · card.json | — |
granite-3.0-1b-a400m-instruct-int8 | text-generation | — | not measured | not measured | galaxy-s26 cpu: fallback · 23.255 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.18.0 · Android 16 (BP4A.251205.006) · 2026-10-21galaxy-s26 gpu: pass · 27.285 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.18.0 · Android 16 (BP4A.251205.006) · 2026-10-21mac-studio-m4-max cpu: pass · 53.05 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.18.0 · macOS 27.0 · 2026-10-21mac-studio-m4-max gpu: pass · 146.34 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.18.0 · macOS 27.0 · 2026-10-21pixel-8a cpu: fallback · 12.8175 tok/s decode stale — latest known 0.19.1Pixel 8a · Google Tensor G3 · litert-lm 0.18.0 · Android 16 (CP1A.260505.005) · 2026-10-21pixel-8a gpu: pass · 12.5825 tok/s decode stale — latest known 0.19.1Pixel 8a · Google Tensor G3 · litert-lm 0.18.0 · Android 16 (CP1A.260505.005) · 2026-10-21 | CARD.md · card.json | — |
granite-4.0-h-1b | text-generation | — | not measured | not measured | galaxy-s26 cpu: fallback · 29.39 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-05galaxy-s26 gpu: pass · 24.06 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-08-24mac-studio-m4-max cpu: pass stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.16.0 · 2026-08-12raspberry-pi-5 cpu: pass · 4.38 tok/s decode stale — latest known 0.19.1Raspberry Pi 5 Model B Rev 1.1 · Broadcom BCM2712 · litert-lm 0.16.1 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-09-01 | CARD.md · card.json | — |
granite-4.0-h-350m-fp16 | text-generation | — | not measured | not measured | galaxy-s26 cpu: fallback · 9.54 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-05galaxy-s26 gpu: load_failed stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-08-24mac-studio-m4-max cpu: pass stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.16.0 · 2026-08-12raspberry-pi-5 cpu: pass · 3.83 tok/s decode stale — latest known 0.19.1Raspberry Pi 5 Model B Rev 1.1 · Broadcom BCM2712 · litert-lm 0.16.1 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-09-02 | CARD.md · card.json | — |
granite-4.0-h-350m-int8 | text-generation | — | not measured | not measured | galaxy-s26 cpu: fallback · 11.48 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-05galaxy-s26 gpu: load_failed stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-08-24mac-studio-m4-max cpu: pass stale — latest known 0.19.1Mac Studio (M4 Max) · litert-lm 0.17.0 · 2026-09-06pixel-8a cpu: fallback · 8.86 tok/s decode stale — latest known 0.19.1Pixel 8a · Tensor G3 · litert-lm 0.16.0 · Android 16 · 2026-09-05raspberry-pi-5 cpu: pass · 5.57 tok/s decode stale — latest known 0.19.1Raspberry Pi 5 Model B Rev 1.1 · Broadcom BCM2712 · litert-lm 0.16.1 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-09-02 | CARD.md · card.json | — |
granite-4.0-h-350m-int8-gpu | text-generation | — | not measured | not measured | galaxy-s26 cpu: pass · 95.16 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-08-27galaxy-s26 gpu: pass · 47.87 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-08-27iphone-17-pro cpu: pass stale — latest known 0.19.1iPhone 17 Pro · A19-Pro · litert-lm 0.15.0 · iOS 27.0 · 2026-08-28iphone-17-pro gpu: pass stale — latest known 0.19.1iPhone 17 Pro · A19-Pro · litert-lm 0.15.0 · iOS 27.0 · 2026-08-28pixel-8a cpu: fallback · 40.7 tok/s decode stale — latest known 0.19.1Pixel 8a · Tensor G3 · litert-lm 0.16.0 · Android 16 · 2026-09-07pixel-8a gpu: pass · 20.7 tok/s decode stale — latest known 0.19.1Pixel 8a · Tensor G3 · litert-lm 0.16.0 · Android 16 · 2026-09-07raspberry-pi-5 cpu: pass · 14.04 tok/s decode stale — latest known 0.19.1Raspberry Pi 5 Model B Rev 1.1 · Broadcom BCM2712 · litert-lm 0.16.1 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-09-02 | CARD.md · card.json | — |
granite-4.0-micro-int4 | text-generation | — | not measured | not measured | galaxy-s26 gpu: pass · 20.2425 tok/s decodeGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm prebuilt-adac974b · Android 16 (BP4A.251205.006) · 2026-10-15mac-studio-m4-max cpu: pass · 23.615 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.18.0 · macOS 27.0 · 2026-10-14mac-studio-m4-max gpu: pass · 111.25 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.18.0 · macOS 27.0 · 2026-10-14pixel-8a cpu: fallback · 3.085 tok/s decodePixel 8a · Tensor G3 · litert-lm prebuilt-adac974b · Android 16 (CP1A.260505.005) · 2026-10-15pixel-8a gpu: pass · 6.975 tok/s decodePixel 8a · Tensor G3 · litert-lm prebuilt-adac974b · Android 16 (CP1A.260505.005) · 2026-10-16 | CARD.md · card.json | — |
granite-4.0-micro-int8 | text-generation | — | not measured | not measured | mac-studio-m4-max cpu: pass · 21.355 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.18.0 · macOS 27.0 · 2026-10-14mac-studio-m4-max gpu: pass · 89.335 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.18.0 · macOS 27.0 · 2026-10-14pixel-8a cpu: fallback · 2.53 tok/s decodePixel 8a · Tensor G3 · litert-lm prebuilt-adac974b · Android 16 (CP1A.260505.005) · 2026-10-15 | CARD.md · card.json | — |
granite-4.1-3b-int4 | text-generation | — | not measured | not measured | galaxy-s26 gpu: pass · 15.28 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-08-24mac-m4-max gpu: pass · 38.06 tok/s decode stale — latest known 0.19.1Mac Studio M4 Max · litert-lm 0.16.0 · 2026-08-17pixel-8a gpu: pass · 7.07 tok/s decode stale — latest known 0.19.1Pixel 8a · Tensor G3 · litert-lm 0.16.0 · 2026-08-17raspberry-pi-5 cpu: pass · 2.16 tok/s decode stale — latest known 0.19.1Raspberry Pi 5 Model B Rev 1.1 · Broadcom BCM2712 · litert-lm 0.16.1 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-09-01 | CARD.md · card.json | — |
granite-4.1-3b-int8 | text-generation | — | not measured | not measured | galaxy-s26 cpu: fallback · 8.61 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-06galaxy-s26 gpu: pass · 9.52 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-08-25mac-m4-max gpu: pass · 32.74 tok/s decode stale — latest known 0.19.1Mac Studio M4 Max · litert-lm 0.16.0 · 2026-08-17pixel-8a gpu: pass · 3.86 tok/s decode stale — latest known 0.19.1Pixel 8a · Tensor G3 · litert-lm 0.16.0 · 2026-08-17raspberry-pi-5 cpu: pass · 1.73 tok/s decode stale — latest known 0.19.1Raspberry Pi 5 Model B Rev 1.1 · Broadcom BCM2712 · litert-lm 0.16.1 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-09-01 | CARD.md · card.json | — |
granite-4.1-8b-int4 | text-generation | — | not measured | not measured | galaxy-s26 gpu: pass · 5.3475 tok/s decodeGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm prebuilt-adac974b · Android 16 (BP4A.251205.006) · 2026-10-15mac-studio-m4-max cpu: pass stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.17.1 · macOS 27.0 (26A428) · 2026-10-14mac-studio-m4-max cpu: pass · 11.68 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.18.0 · macOS 27.0 (26A428) · 2026-10-15mac-studio-m4-max gpu: pass · 58.305 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.18.0 · macOS 27.0 (26A428) · 2026-10-15pixel-8a gpu: load_failedPixel 8a · Tensor G3 · litert-lm prebuilt-adac974b · Android 16 (CP1A.260505.005) · 2026-10-14 | CARD.md · card.json | — |
granite-4.1-8b-int8 | text-generation | — | not measured | not measured | mac-studio-m4-max cpu: pass stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.17.1 · macOS 27.0 (26A428) · 2026-10-14mac-studio-m4-max cpu: pass · 9.815 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.18.0 · macOS 27.0 (26A428) · 2026-10-15mac-studio-m4-max gpu: pass · 46.07 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.18.0 · macOS 27.0 (26A428) · 2026-10-15 | CARD.md · card.json | — |
granite-4.2-3b-int4 | text-generation | — | not measured | not measured | galaxy-s26 cpu: fallback · 9.83 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-08-31galaxy-s26 gpu: pass · 15.61 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-08-31iphone-17-pro cpu: pass stale — latest known 0.19.1iPhone 17 Pro · A19-Pro · litert-lm 0.15.0 · iOS 27.0 · 2026-08-31iphone-17-pro gpu: pass stale — latest known 0.19.1iPhone 17 Pro · A19-Pro · litert-lm 0.15.0 · iOS 27.0 · 2026-08-31mac-studio-m4-max cpu: pass · 21.47 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · litert-lm 0.16.0 · 2026-08-31mac-studio-m4-max gpu: pass · 85.46 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · litert-lm 0.16.0 · 2026-08-31raspberry-pi-5 cpu: pass · 2.15 tok/s decode stale — latest known 0.19.1Raspberry Pi 5 Model B Rev 1.1 · Broadcom BCM2712 · litert-lm 0.16.1 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-09-02 | CARD.md · card.json | — |
granite-4.2-3b-int8 | text-generation | — | not measured | not measured | galaxy-s26 cpu: fallback · 7.03 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-08-31galaxy-s26 gpu: pass · 9.22 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-08-31mac-studio-m4-max cpu: pass · 21.12 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · litert-lm 0.16.0 · 2026-08-31mac-studio-m4-max gpu: pass · 70.65 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · litert-lm 0.16.0 · 2026-08-31raspberry-pi-5 cpu: pass · 1.73 tok/s decode stale — latest known 0.19.1Raspberry Pi 5 Model B Rev 1.1 · Broadcom BCM2712 · litert-lm 0.16.1 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-09-02 | CARD.md · card.json | — |
granite-docling-258m | image-text-to-text | — | not measured | not measured | galaxy-s26 cpu: fallback stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.1 · Android 16 · 2026-08-24galaxy-s26 cpu: fallback · 33.49 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-05galaxy-s26 gpu: load_failed stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.1 · Android 16 · 2026-08-24pixel-8a cpu: fallback · 23.25 tok/s decode stale — latest known 0.19.1Pixel 8a · Tensor G3 · litert-lm 0.16.0 · Android 16 · 2026-09-05pixel-8a gpu: run_failed stale — latest known 0.19.1Pixel 8a · Tensor G3 · litert-lm 0.16.0 · Android 16 · 2026-09-05 | CARD.md · card.json | — |
granite-speech-5.0-2b | automatic-speech-recognition | — | not measured | not measured | galaxy-s26 cpu: fallbackGalaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU) · litert-lm 0.19.1 · Android 16 (BP4A.251205.006) · 2026-11-07galaxy-s26 cpu: fallbackGalaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU) · litert-lm 0.19.1 · Android 16 (BP4A.251205.006) · 2026-11-06galaxy-s26 cpu: fallbackGalaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU) · litert-lm 0.19.1 · Android 16 (BP4A.251205.006) · 2026-11-06galaxy-s26 cpu: fallback · 10.0332 tok/s decodeGalaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU) · litert-lm 0.19.1 · Android 16 (BP4A.251205.006) · 2026-11-07galaxy-s26 gpu: fallbackGalaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU) · litert-lm 0.19.1 · Android 16 (BP4A.251205.006) · 2026-11-07galaxy-s26 gpu: fallbackGalaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU) · litert-lm 0.19.1 · Android 16 (BP4A.251205.006) · 2026-11-06galaxy-s26 gpu: fallbackGalaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU) · litert-lm 0.19.1 · Android 16 (BP4A.251205.006) · 2026-11-06galaxy-s26 gpu: fallback · 14.8803 tok/s decodeGalaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU) · litert-lm 0.19.1 · Android 16 (BP4A.251205.006) · 2026-11-07mac-studio-m4-max cpu: passMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.1 · macOS 27.0 (26A428) · 2026-11-06mac-studio-m4-max cpu: passMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.1 · macOS 27.0 (26A428) · 2026-11-06mac-studio-m4-max cpu: pass · 26.65 tok/s decodeMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.1 · macOS 27.0 (26A428) · 2026-11-06mac-studio-m4-max gpu: passMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.1 · macOS 27.0 (26A428) · 2026-11-06mac-studio-m4-max gpu: passMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.1 · macOS 27.0 (26A428) · 2026-11-06mac-studio-m4-max gpu: load_failedMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.1 · macOS 27.0 (26A428) · 2026-11-06mac-studio-m4-max gpu: pass · 84.18 tok/s decodeMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.1 · macOS 27.0 (26A428) · 2026-11-06pixel-11-pro cpu: fallbackPixel 11 Pro · Google Tensor G6 · litert-lm 0.19.1 · Android 17 (CD1A.260905.001.B1) · 2026-11-07pixel-11-pro cpu: fallbackPixel 11 Pro · Google Tensor G6 · litert-lm 0.19.1 · Android 17 (CD1A.260905.001.B1) · 2026-11-06pixel-11-pro cpu: fallback · 4.7182 tok/s decodePixel 11 Pro · Google Tensor G6 · litert-lm 0.19.1 · Android 17 (CD1A.260905.001.B1) · 2026-11-06pixel-11-pro gpu: fallbackPixel 11 Pro · Google Tensor G6 · litert-lm 0.19.1 · Android 17 (CD1A.260905.001.B1) · 2026-11-07pixel-11-pro gpu: fallbackPixel 11 Pro · Google Tensor G6 · litert-lm 0.19.1 · Android 17 (CD1A.260905.001.B1) · 2026-11-06pixel-11-pro gpu: fallback · 7.6361 tok/s decodePixel 11 Pro · Google Tensor G6 · litert-lm 0.19.1 · Android 17 (CD1A.260905.001.B1) · 2026-11-07 | CARD.md · card.json | — |
granite-speech-5.0-2b-ekv2048 | automatic-speech-recognition | — | not measured | not measured | galaxy-s26 cpu: fallbackGalaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU) · litert-lm 0.19.1 · Android 16 (BP4A.251205.006) · 2026-11-07galaxy-s26 cpu: fallbackGalaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU) · litert-lm 0.19.1 · Android 16 (BP4A.251205.006) · 2026-11-07galaxy-s26 cpu: fallbackGalaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU) · litert-lm 0.19.1 · Android 16 (BP4A.251205.006) · 2026-11-07galaxy-s26 cpu: fallbackGalaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU) · litert-lm 0.19.1 · Android 16 (BP4A.251205.006) · 2026-11-07galaxy-s26 cpu: fallbackGalaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU) · litert-lm 0.19.1 · Android 16 (BP4A.251205.006) · 2026-11-07galaxy-s26 cpu: fallbackGalaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU) · litert-lm 0.19.1 · Android 16 (BP4A.251205.006) · 2026-11-07galaxy-s26 cpu: fallbackGalaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU) · litert-lm 0.19.1 · Android 16 (BP4A.251205.006) · 2026-11-07galaxy-s26 cpu: fallback · 15.3363 tok/s decodeGalaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU) · litert-lm 0.19.1 · Android 16 (BP4A.251205.006) · 2026-11-07galaxy-s26 gpu: fallbackGalaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU) · litert-lm 0.19.1 · Android 16 (BP4A.251205.006) · 2026-11-07galaxy-s26 gpu: fallbackGalaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU) · litert-lm 0.19.1 · Android 16 (BP4A.251205.006) · 2026-11-07galaxy-s26 gpu: fallbackGalaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU) · litert-lm 0.19.1 · Android 16 (BP4A.251205.006) · 2026-11-07mac-studio-m4-max cpu: passMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.1 · macOS 27.0 (26A428) · 2026-11-07mac-studio-m4-max cpu: passMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.1 · macOS 27.0 (26A428) · 2026-11-07mac-studio-m4-max cpu: pass · 36.21 tok/s decodeMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.1 · macOS 27.0 (26A428) · 2026-11-07mac-studio-m4-max gpu: passMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.1 · macOS 27.0 (26A428) · 2026-11-07mac-studio-m4-max gpu: passMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.1 · macOS 27.0 (26A428) · 2026-11-07mac-studio-m4-max gpu: pass · 109.87 tok/s decodeMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.1 · macOS 27.0 (26A428) · 2026-11-07pixel-11-pro cpu: fallbackPixel 11 Pro · Google Tensor G6 · litert-lm 0.19.1 · Android 17 (CD1A.260905.001.B1) · 2026-11-07pixel-11-pro gpu: fallbackPixel 11 Pro · Google Tensor G6 · litert-lm 0.19.1 · Android 17 (CD1A.260905.001.B1) · 2026-11-07 | CARD.md · card.json | — |
granite4h-1b-gpufix | text-generation | — | not measured | not measured | pixel-8a cpu: pass · 4.34 tok/s decode stale — latest known 0.19.1pixel-8a · Google Tensor G3 · litert-lm 0.16.0 · 2026-08-13pixel-8a gpu: pass · 8.81 tok/s decode stale — latest known 0.19.1pixel-8a · Google Tensor G3 · litert-lm 0.16.0 · 2026-08-13 | CARD.md · card.json | — |
hsemotion-b0 | image-classification | — | pass · 72.14 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | pass · 2.565 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | raspberry-pi-5 cpu_xnnpack: pass · 24.767 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
hy-mt2-1.8b-int8 | text-generation | — | not measured | not measured | galaxy-s26 cpu: fallback · 12.27 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-05galaxy-s26 cpu: fallback · 14.37 tok/s decodeGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm prebuilt-adac974b · Android 16 · 2026-10-05galaxy-s26 gpu: pass · 25.62 tok/s decodeGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm prebuilt-adac974b · Android 16 · 2026-10-05galaxy-s26 gpu: pass · 20.84 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-05galaxy-s26 gpu: pass · 27.38 tok/s decodeGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm prebuilt-adac974b · Android 16 · 2026-10-05mac-studio-m4-max cpu: fallback · 16.91 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.17.1 · 2026-10-05mac-studio-m4-max cpu: fallback · 31.9 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.17.1 · 2026-10-05mac-studio-m4-max gpu: pass · 137.15 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.17.1 · 2026-10-05mac-studio-m4-max gpu: pass · 150.72 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.17.1 · 2026-10-05pixel-8a cpu: fallback · 8.3 tok/s decode stale — latest known 0.19.1Pixel 8a · Tensor G3 · litert-lm 0.16.0 · Android 16 · 2026-09-05raspberry-pi-5 cpu: pass · 2.66 tok/s decode stale — latest known 0.19.1Raspberry Pi 5 Model B Rev 1.1 · Broadcom BCM2712 · litert-lm 0.16.1 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-09-02 | CARD.md · card.json | — |
internvl3-1b | image-text-to-text | — | not measured | not measured | galaxy-s26 cpu: fallback · 81.7 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-05galaxy-s26 cpu: pass · 68.14 tok/s decodeGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm selfbuilt-1dadd00c · Android 16 · 2026-09-26galaxy-s26 gpu: load_failedGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm selfbuilt-1dadd00c · Android 16 · 2026-09-26galaxy-s26 gpu: pass · 85.42 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-08-25mac-studio-m4-max cpu: pass · 78.7 tok/s decodeMac Studio (M4 Max) · Apple M4 Max · litert-lm selfbuilt-1dadd00c · macOS 27.0 (26A428) · 2026-10-01mac-studio-m4-max gpu: load_failedMac Studio (M4 Max) · Apple M4 Max · litert-lm selfbuilt-1dadd00c · macOS 27.0 (26A428) · 2026-10-01 | CARD.md · card.json | — |
internvl3-2b | image-text-to-text | — | not measured | not measured | galaxy-s26 cpu: fallback · 43.67 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-05galaxy-s26 gpu: pass · 34.4 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-08-25 | CARD.md · card.json | — |
internvl3_5-1b | image-text-to-text | — | not measured | not measured | galaxy-s26 cpu: fallback · 26.13 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-05galaxy-s26 gpu: pass · 43.45 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-08-25 | CARD.md · card.json | — |
jan-nano | text-generation | — | not measured | not measured | mac-studio-m4-max gpu: pass · 68.98 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.14.0 · 2026-07-23raspberry-pi-5 cpu: pass · 1.5 tok/s decode stale — latest known 0.19.1Raspberry Pi 5 Model B Rev 1.1 · Broadcom BCM2712 · litert-lm 0.16.1 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-09-01 | CARD.md · card.json | — |
jan-nano-model-block32 | text-generation | — | not measured | not measured | galaxy-s26 cpu: fallback · 5.94 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-05galaxy-s26 gpu: pass · 9.3 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-08-24raspberry-pi-5 cpu: pass · 1.45 tok/s decode stale — latest known 0.19.1Raspberry Pi 5 Model B Rev 1.1 · Broadcom BCM2712 · litert-lm 0.16.1 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-09-01 | CARD.md · card.json | — |
japanese-zipformer-base | automatic-speech-recognition | — | run_failed stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-20 | pass · 61.167 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-20 | raspberry-pi-5 cpu_xnnpack: pass · 2641.607 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
julia-1__julia1_s1024_fp32 | text-classification | gpu_mldrift @ litert 2.2.0: 97.1% · 51 partition(s) | pass · 2014.363 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-10-09 | pass · 43.43 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-10-09 | galaxy-s26 gpu_mldrift: pass · 298.463958 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (SDK 36) · 2026-09-30 | CARD.md · card.json | run it |
julia-1__julia1_s512_fp32 | text-classification | gpu_mldrift @ litert 2.2.0: 97.1% · 51 partition(s) | pass · 845.915 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-10-09 | pass · 23.355 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-10-09 | galaxy-s26 gpu_mldrift: output_mismatch · 80.711094 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (SDK 36) · 2026-09-30galaxy-s26 gpu_mldrift: output_mismatch · 112.052083 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (SDK 36) · 2026-09-30galaxy-s26 npu_qnn: output_mismatch · 29.283906 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (SDK 36) · 2026-09-30 | CARD.md · card.json | run it |
kimi-linear-48b-a3b-int8-ekv4096 | text-generation | — | not measured | not measured | mac-studio-m4-max cpu: run_failedMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.1 · macOS 27.0 (26A428) · 2026-11-15mac-studio-m4-max cpu: run_failedMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.2 · macOS 27.0 (26A428) · 2026-11-15mac-studio-m4-max gpu: load_failedMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.1 · macOS 27.0 (26A428) · 2026-11-15mac-studio-m4-max gpu: load_failedMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.2 · macOS 27.0 (26A428) · 2026-11-15 | CARD.md · card.json | — |
kimi-linear-48b-a3b-mixed-int8-cpu-ekv4096 | text-generation | — | not measured | not measured | mac-studio-m4-max cpu: run_failedMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.1 · macOS 27.0 (26A428) · 2026-11-13mac-studio-m4-max cpu: run_failedMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.2 · macOS 27.0 (26A428) · 2026-11-13mac-studio-m4-max gpu: load_failedMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.1 · macOS 27.0 (26A428) · 2026-11-13mac-studio-m4-max gpu: load_failedMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.2 · macOS 27.0 (26A428) · 2026-11-13 | CARD.md · card.json | — |
l2cs-gaze360 | image-classification | — | pass · 1997.273 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | pass · 11.758 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | galaxy-s26 gpu_mldrift: pass · 11.052 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 · 2026-08-26galaxy-s26 npu_qnn: pass · 3.591 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · QAIRT (Hexagon, JIT on-device) · Android 16 · 2026-08-26raspberry-pi-5 cpu_xnnpack: pass · 349.393 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
lfm2-2.6b-exp-int4 | text-generation | — | not measured | not measured | galaxy-s26 cpu: fallback · 21.215 tok/s decodeGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm prebuilt-adac974b · Android 16 (BP4A.251205.006) · 2026-10-12galaxy-s26 gpu: pass · 27.615 tok/s decodeGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm prebuilt-adac974b · Android 16 (BP4A.251205.006) · 2026-10-12mac-studio-m4-max cpu: pass · 48.04 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.18.0 · macOS 27.0 · 2026-10-12mac-studio-m4-max gpu: pass · 161.77 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.18.0 · macOS 27.0 · 2026-10-12pixel-8a cpu: fallback · 8.005 tok/s decodePixel 8a · Tensor G3 · litert-lm prebuilt-adac974b · Android 16 (CP1A.260505.005) · 2026-10-12pixel-8a gpu: pass · 15.9 tok/s decodePixel 8a · Tensor G3 · litert-lm prebuilt-adac974b · Android 16 (CP1A.260505.005) · 2026-10-12 | CARD.md · card.json | — |
lfm2-2.6b-exp-int8 | text-generation | — | not measured | not measured | galaxy-s26 cpu: fallback · 18.545 tok/s decodeGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm prebuilt-adac974b · Android 16 (BP4A.251205.006) · 2026-10-12galaxy-s26 gpu: pass · 20.435 tok/s decodeGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm prebuilt-adac974b · Android 16 (BP4A.251205.006) · 2026-10-12mac-studio-m4-max cpu: run_failed stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.17.1 · macOS 27.0 · 2026-10-13mac-studio-m4-max cpu: pass · 38.27 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.18.0 · macOS 27.0 · 2026-10-12mac-studio-m4-max gpu: pass · 123.39 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.18.0 · macOS 27.0 · 2026-10-12pixel-8a cpu: fallback · 5.845 tok/s decodePixel 8a · Tensor G3 · litert-lm prebuilt-adac974b · Android 16 (CP1A.260505.005) · 2026-10-12pixel-8a gpu: pass · 11.095 tok/s decodePixel 8a · Tensor G3 · litert-lm prebuilt-adac974b · Android 16 (CP1A.260505.005) · 2026-10-12 | CARD.md · card.json | — |
lfm2-2.6b-int4 | text-generation | — | not measured | not measured | galaxy-s26 cpu: fallback · 22.85 tok/s decodeGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm prebuilt-adac974b · Android 16 (BP4A.251205.006) · 2026-10-12galaxy-s26 gpu: pass · 27.84 tok/s decodeGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm prebuilt-adac974b · Android 16 (BP4A.251205.006) · 2026-10-12mac-studio-m4-max cpu: pass · 49.07 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.18.0 · macOS 27.0 · 2026-10-12mac-studio-m4-max gpu: pass · 163.77 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.18.0 · macOS 27.0 · 2026-10-12pixel-8a cpu: fallback · 7.375 tok/s decodePixel 8a · Tensor G3 · litert-lm prebuilt-adac974b · Android 16 (CP1A.260505.005) · 2026-10-12pixel-8a gpu: pass · 15.845 tok/s decodePixel 8a · Tensor G3 · litert-lm prebuilt-adac974b · Android 16 (CP1A.260505.005) · 2026-10-12 | CARD.md · card.json | — |
lfm2-2.6b-int8 | text-generation | — | not measured | not measured | galaxy-s26 cpu: fallback · 17.07 tok/s decodeGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm prebuilt-adac974b · Android 16 (BP4A.251205.006) · 2026-10-12galaxy-s26 gpu: pass · 20.39 tok/s decodeGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm prebuilt-adac974b · Android 16 (BP4A.251205.006) · 2026-10-12mac-studio-m4-max cpu: run_failed stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.17.1 · macOS 27.0 · 2026-10-12mac-studio-m4-max cpu: pass · 39.18 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.18.0 · macOS 27.0 · 2026-10-12mac-studio-m4-max gpu: pass · 125.63 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.18.0 · macOS 27.0 · 2026-10-12pixel-8a cpu: fallback · 6.655 tok/s decodePixel 8a · Tensor G3 · litert-lm prebuilt-adac974b · Android 16 (CP1A.260505.005) · 2026-10-12pixel-8a gpu: pass · 11.105 tok/s decodePixel 8a · Tensor G3 · litert-lm prebuilt-adac974b · Android 16 (CP1A.260505.005) · 2026-10-12 | CARD.md · card.json | — |
lfm2-700m-int4 | text-generation | — | not measured | not measured | galaxy-s26 cpu: fallback · 57.12 tok/s decodeGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm prebuilt-adac974b · Android 16 · 2026-10-12galaxy-s26 gpu: pass · 73.25 tok/s decodeGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm prebuilt-adac974b · Android 16 · 2026-10-12mac-studio-m4-max cpu: pass · 105.965 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.18.0 · macOS 27.0 · 2026-10-12mac-studio-m4-max gpu: pass · 348.26 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.18.0 · macOS 27.0 · 2026-10-12pixel-8a cpu: fallback · 30.325 tok/s decodePixel 8a · Tensor G3 · litert-lm prebuilt-adac974b · Android 16 · 2026-10-12pixel-8a gpu: pass · 39.905 tok/s decodePixel 8a · Tensor G3 · litert-lm prebuilt-adac974b · Android 16 · 2026-10-12 | CARD.md · card.json | — |
lfm2-700m-int8 | text-generation | — | not measured | not measured | galaxy-s26 cpu: fallback · 45.65 tok/s decodeGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm prebuilt-adac974b · Android 16 · 2026-10-12galaxy-s26 gpu: pass · 58.965 tok/s decodeGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm prebuilt-adac974b · Android 16 · 2026-10-12mac-studio-m4-max cpu: run_failed stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.17.1 · macOS 27.0 · 2026-10-12mac-studio-m4-max cpu: pass · 90.87 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.18.0 · macOS 27.0 · 2026-10-12mac-studio-m4-max gpu: pass · 306.34 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.18.0 · macOS 27.0 · 2026-10-12pixel-8a cpu: fallback · 22.725 tok/s decodePixel 8a · Tensor G3 · litert-lm prebuilt-adac974b · Android 16 · 2026-10-12pixel-8a gpu: pass · 34.75 tok/s decodePixel 8a · Tensor G3 · litert-lm prebuilt-adac974b · Android 16 · 2026-10-12 | CARD.md · card.json | — |
lfm2.5-1.2b-instruct-int4 | text-generation | — | not measured | not measured | galaxy-s26 gpu: load_failed stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-08-23mac-m4-max gpu: load_failed stale — latest known 0.19.1Mac Studio (Mac16,9), Apple M4 Max · litert-lm 0.18.0 · macOS 27.0 · 2026-10-07mac-studio-m4-max cpu: pass stale — latest known 0.19.1Mac Studio (M4 Max) · litert-lm 0.17.0 · 2026-09-06mac-studio-m4-max gpu: load_failed stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.14.0 · 2026-07-22pixel-8a gpu: load_failed stale — latest known 0.19.1Pixel 8a · Google Tensor G3 · litert-lm 0.16.0 · 2026-08-12raspberry-pi-5 cpu: pass · 9.29 tok/s decode stale — latest known 0.19.1Raspberry Pi 5 Model B Rev 1.1 · Broadcom BCM2712 · litert-lm 0.16.1 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-09-01 | CARD.md · card.json | — |
lfm2.5-1.2b-instruct-int8 | text-generation | — | not measured | not measured | galaxy-s26 cpu: fallback · 40.96 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-05mac-studio-m4-max cpu: pass stale — latest known 0.19.1Mac Studio (M4 Max) · litert-lm 0.17.0 · 2026-09-06pixel-8a cpu: fallback · 20.4 tok/s decode stale — latest known 0.19.1Pixel 8a · Tensor G3 · litert-lm 0.16.0 · Android 16 · 2026-09-05raspberry-pi-5 cpu: pass · 7.02 tok/s decode stale — latest known 0.19.1Raspberry Pi 5 Model B Rev 1.1 · Broadcom BCM2712 · litert-lm 0.16.1 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-09-01 | CARD.md · card.json | — |
lfm2.5-1.2b-jp-int4 | text-generation | — | not measured | not measured | galaxy-s26 gpu: load_failed stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-08-24mac-studio-m4-max cpu: pass stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.16.0 · 2026-08-12raspberry-pi-5 cpu: pass · 9.22 tok/s decode stale — latest known 0.19.1Raspberry Pi 5 Model B Rev 1.1 · Broadcom BCM2712 · litert-lm 0.16.1 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-09-01 | CARD.md · card.json | — |
lfm2.5-1.2b-jp-int8 | text-generation | — | not measured | not measured | galaxy-s26 cpu: fallback · 41.32 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-05galaxy-s26 gpu: load_failed stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-08-24mac-studio-m4-max cpu: pass stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.16.0 · 2026-08-12raspberry-pi-5 cpu: pass · 7.03 tok/s decode stale — latest known 0.19.1Raspberry Pi 5 Model B Rev 1.1 · Broadcom BCM2712 · litert-lm 0.16.1 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-09-01 | CARD.md · card.json | — |
lfm2.5-1.2b-thinking-int4 | text-generation | — | not measured | not measured | galaxy-s26 gpu: load_failed stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-08-24mac-studio-m4-max cpu: pass stale — latest known 0.19.1Mac Studio (M4 Max) · litert-lm 0.17.0 · 2026-09-06raspberry-pi-5 cpu: pass · 9.25 tok/s decode stale — latest known 0.19.1Raspberry Pi 5 Model B Rev 1.1 · Broadcom BCM2712 · litert-lm 0.16.1 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-09-01 | CARD.md · card.json | — |
lfm2.5-1.2b-thinking-int8 | text-generation | — | not measured | not measured | galaxy-s26 cpu: fallback · 35.55 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-05galaxy-s26 gpu: load_failed stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-08-24mac-studio-m4-max cpu: pass stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.16.0 · 2026-08-12raspberry-pi-5 cpu: pass · 7.07 tok/s decode stale — latest known 0.19.1Raspberry Pi 5 Model B Rev 1.1 · Broadcom BCM2712 · litert-lm 0.16.1 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-09-01 | CARD.md · card.json | — |
lfm2.5-2.6b-int4 | text-generation | — | not measured | not measured | galaxy-s26 gpu: pass · 25.31 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-08-24raspberry-pi-5 cpu: pass · 4.91 tok/s decode stale — latest known 0.19.1Raspberry Pi 5 Model B Rev 1.1 · Broadcom BCM2712 · litert-lm 0.16.1 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-09-01 | CARD.md · card.json | — |
lfm2.5-2.6b-int8 | text-generation | — | not measured | not measured | galaxy-s26 cpu: fallback · 15.53 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-06galaxy-s26 gpu: pass · 19.72 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-08-25raspberry-pi-5 cpu: pass · 3.5 tok/s decode stale — latest known 0.19.1Raspberry Pi 5 Model B Rev 1.1 · Broadcom BCM2712 · litert-lm 0.16.1 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-09-01 | CARD.md · card.json | — |
lfm2.5-230m-int4 | text-generation | — | not measured | not measured | galaxy-s26 cpu: fallback · 50.68 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-01galaxy-s26 gpu: pass · 41.62 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-01iphone-17-pro cpu: pass · 93.9 tok/s decode stale — latest known 0.19.1iPhone 17 Pro · A19-Pro · litert-lm 0.16.0 · iOS 27.0 · 2026-09-01iphone-17-pro gpu: pass · 161.7 tok/s decode stale — latest known 0.19.1iPhone 17 Pro · A19-Pro · litert-lm 0.16.0 · iOS 27.0 · 2026-09-01mac-studio-m4-max cpu: pass · 149.35 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · litert-lm 0.16.0 · 2026-09-01mac-studio-m4-max gpu: pass · 555.27 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · litert-lm 0.16.0 · 2026-09-01 | CARD.md · card.json | — |
lfm2.5-230m-int8 | text-generation | — | not measured | not measured | galaxy-s26 cpu: fallback · 94.69 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-01galaxy-s26 gpu: pass · 121.69 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-01iphone-17-pro cpu: pass · 86.55 tok/s decode stale — latest known 0.19.1iPhone 17 Pro · A19-Pro · litert-lm 0.16.0 · iOS 27.0 · 2026-09-01iphone-17-pro gpu: pass · 143.19 tok/s decode stale — latest known 0.19.1iPhone 17 Pro · A19-Pro · litert-lm 0.16.0 · iOS 27.0 · 2026-09-01mac-studio-m4-max cpu: pass · 153.82 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · litert-lm 0.16.0 · 2026-09-01mac-studio-m4-max gpu: pass · 561.24 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · litert-lm 0.16.0 · 2026-09-01pixel-8a cpu: fallback · 58.66 tok/s decode stale — latest known 0.19.1Pixel 8a · Tensor G3 · litert-lm 0.16.0 · Android 16 · 2026-09-05pixel-8a gpu: pass · 39.39 tok/s decode stale — latest known 0.19.1Pixel 8a · Tensor G3 · litert-lm 0.16.0 · Android 16 · 2026-09-05 | CARD.md · card.json | — |
lfm2.5-350m-int4 | text-generation | — | not measured | not measured | galaxy-s26 cpu: fallback · 90.63 tok/s decodeGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm prebuilt-adac974b · Android 16 · 2026-10-12galaxy-s26 gpu: pass · 106.86 tok/s decodeGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm prebuilt-adac974b · Android 16 · 2026-10-12mac-studio-m4-max cpu: pass · 123.11 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.18.0 · macOS 27.0 · 2026-10-12mac-studio-m4-max gpu: pass · 425.85 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.18.0 · macOS 27.0 · 2026-10-12pixel-8a cpu: fallback · 51.405 tok/s decodePixel 8a · Tensor G3 · litert-lm prebuilt-adac974b · Android 16 · 2026-10-12pixel-8a gpu: pass · 50.055 tok/s decodePixel 8a · Tensor G3 · litert-lm prebuilt-adac974b · Android 16 · 2026-10-12 | CARD.md · card.json | — |
lfm2.5-350m-int8 | text-generation | — | not measured | not measured | galaxy-s26 cpu: fallback · 80.31 tok/s decodeGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm prebuilt-adac974b · Android 16 · 2026-10-12galaxy-s26 gpu: pass · 92.895 tok/s decodeGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm prebuilt-adac974b · Android 16 · 2026-10-12mac-studio-m4-max cpu: run_failed stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.17.1 · macOS 27.0 · 2026-10-12mac-studio-m4-max cpu: pass · 134 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.18.0 · macOS 27.0 · 2026-10-12mac-studio-m4-max gpu: pass · 423.73 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.18.0 · macOS 27.0 · 2026-10-12pixel-8a cpu: fallback · 41.81 tok/s decodePixel 8a · Tensor G3 · litert-lm prebuilt-adac974b · Android 16 · 2026-10-12pixel-8a gpu: pass · 47.895 tok/s decodePixel 8a · Tensor G3 · litert-lm prebuilt-adac974b · Android 16 · 2026-10-12 | CARD.md · card.json | — |
lfm2.5-vl-3b-int8 | image-text-to-text | — | not measured | not measured | galaxy-s26 cpu: fallback · 19.87 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-05galaxy-s26 gpu: pass · 20.15 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-08-25 | CARD.md · card.json | — |
lfm25-12b-instruct-int4-gpu-093 | text-generation | — | not measured | not measured | galaxy-s26 cpu: fallback · 24.09 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-08-23galaxy-s26 cpu: fallback · 40.7 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-05galaxy-s26 gpu: pass · 24.74 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-08-23galaxy-s26 gpu: pass · 54.56 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-05pixel-8a gpu: pass · 19.88 tok/s decode stale — latest known 0.19.1Pixel 8a · Google Tensor G3 · litert-lm 0.16.0 · 2026-08-12raspberry-pi-5 cpu: pass · 9.3 tok/s decode stale — latest known 0.19.1Raspberry Pi 5 Model B Rev 1.1 · Broadcom BCM2712 · litert-lm 0.16.1 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-09-01 | CARD.md · card.json | — |
lfm25-12b-int4-gpu | text-generation | — | not measured | not measured | iphone-17-pro gpu: pass · 69.53 tok/s decode stale — latest known 0.19.1iPhone 17 Pro · A19-Pro · litert-lm 0.16.0 · iOS 27.0 · 2026-08-26mac-studio-m4-max gpu: pass · 325.76 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.16.0 · iOS Version 27.0 (Build 26A5416b) · 2026-08-27mac-studio-m4-max gpu: pass · 346.72 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.16.0 · 2026-08-12 | CARD.md · card.json | — |
lfm25-12b-jp-int4-gpu-093 | text-generation | — | not measured | not measured | galaxy-s26 gpu: pass · 54.63 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-08-24iphone-17-pro gpu: pass · 70.02 tok/s decode stale — latest known 0.19.1iphone-17-pro · A19-Pro · litert-lm 0.16.0 · iOS 27.0 · 2026-08-25pixel-8a gpu: pass · 20.48 tok/s decode stale — latest known 0.19.1Pixel 8a · Google Tensor G3 · litert-lm 0.16.0 · 2026-08-12raspberry-pi-5 cpu: pass · 9.24 tok/s decode stale — latest known 0.19.1Raspberry Pi 5 Model B Rev 1.1 · Broadcom BCM2712 · litert-lm 0.16.1 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-09-01 | CARD.md · card.json | — |
lfm25-12b-jp-int8-gpu | text-generation | — | not measured | not measured | galaxy-s26 gpu: pass · 41.86 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-08-25raspberry-pi-5 cpu: pass · 7.01 tok/s decode stale — latest known 0.19.1Raspberry Pi 5 Model B Rev 1.1 · Broadcom BCM2712 · litert-lm 0.16.1 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-09-01 | CARD.md · card.json | — |
lfm25-12b-thinking-int4-gpu-093 | text-generation | — | not measured | not measured | galaxy-s26 cpu: fallback · 38.23 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-05galaxy-s26 gpu: pass · 48.14 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-08-24iphone-17-pro gpu: pass · 69.66 tok/s decode stale — latest known 0.19.1iphone-17-pro · A19-Pro · litert-lm 0.16.0 · iOS 27.0 · 2026-08-25pixel-8a gpu: pass · 21.48 tok/s decode stale — latest known 0.19.1Pixel 8a · Google Tensor G3 · litert-lm 0.16.0 · 2026-08-12raspberry-pi-5 cpu: pass · 9.29 tok/s decode stale — latest known 0.19.1Raspberry Pi 5 Model B Rev 1.1 · Broadcom BCM2712 · litert-lm 0.16.1 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-09-01 | CARD.md · card.json | — |
lfm25-12b-thinking-int8-gpu | text-generation | — | not measured | not measured | galaxy-s26 gpu: pass · 28.83 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-08-25raspberry-pi-5 cpu: pass · 7.03 tok/s decode stale — latest known 0.19.1Raspberry Pi 5 Model B Rev 1.1 · Broadcom BCM2712 · litert-lm 0.16.1 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-09-01 | CARD.md · card.json | — |
lfm25-12b-wi8-composite | text-generation | — | not measured | not measured | mac-studio-m4-max gpu: pass · 262.55 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.16.0 · 2026-08-12 | CARD.md · card.json | — |
lfm25-12b-wi8-gpu | text-generation | — | not measured | not measured | mac-studio-m4-max gpu: pass · 272 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.16.0 · 2026-08-12 | CARD.md · card.json | — |
lfm25-encoder-230m | feature-extraction | — | not measured | not measured | not measured | CARD.md · card.json | — |
lfm25-vl-1.6b-int4 | image-text-to-text | — | not measured | not measured | galaxy-s26 cpu: fallback · 50.85 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-05galaxy-s26 gpu: pass · 54.1 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-08-25pixel-8a cpu: pass · 13.64 tok/s decode stale — latest known 0.19.1pixel-8a · Google Tensor G3 · litert-lm 0.16.0 · 2026-08-13pixel-8a gpu: pass · 22.08 tok/s decode stale — latest known 0.19.1pixel-8a · Google Tensor G3 · litert-lm 0.16.0 · 2026-08-13 | CARD.md · card.json | — |
lfm25-vl-1.6b-int8 | image-text-to-text | — | not measured | not measured | galaxy-s26 cpu: fallback · 41.66 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-05galaxy-s26 gpu: pass · 42.27 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-08-25pixel-8a cpu: pass · 8.38 tok/s decode stale — latest known 0.19.1pixel-8a · Google Tensor G3 · litert-lm 0.16.0 · 2026-08-13pixel-8a gpu: pass · 15.91 tok/s decode stale — latest known 0.19.1pixel-8a · Google Tensor G3 · litert-lm 0.16.0 · 2026-08-13 | CARD.md · card.json | — |
lfm25-vl-3b-int4 | image-text-to-text | — | not measured | not measured | galaxy-s26 gpu: pass · 27.04 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-08-25pixel-8a cpu: pass · 5.77 tok/s decode stale — latest known 0.19.1pixel-8a · Google Tensor G3 · litert-lm 0.16.0 · 2026-08-13pixel-8a gpu: pass · 10.79 tok/s decode stale — latest known 0.19.1pixel-8a · Google Tensor G3 · litert-lm 0.16.0 · 2026-08-13 | CARD.md · card.json | — |
lfm25-vl-450m-int4 | image-text-to-text | — | not measured | not measured | galaxy-s26 cpu: fallback · 99.25 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-05galaxy-s26 gpu: pass · 103.02 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-08-25pixel-8a cpu: pass · 22.62 tok/s decode stale — latest known 0.19.1pixel-8a · Google Tensor G3 · litert-lm 0.16.0 · 2026-08-13pixel-8a gpu: pass · 33.05 tok/s decode stale — latest known 0.19.1pixel-8a · Google Tensor G3 · litert-lm 0.16.0 · 2026-08-13 | CARD.md · card.json | — |
lfm25-vl-450m-int8 | image-text-to-text | — | not measured | not measured | galaxy-s26 cpu: fallback · 70.53 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-05galaxy-s26 gpu: pass · 97.57 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-08-25pixel-8a cpu: pass · 15.75 tok/s decode stale — latest known 0.19.1pixel-8a · Google Tensor G3 · litert-lm 0.16.0 · 2026-08-13pixel-8a gpu: pass · 31.08 tok/s decode stale — latest known 0.19.1pixel-8a · Google Tensor G3 · litert-lm 0.16.0 · 2026-08-13 | CARD.md · card.json | — |
llava-onevision-0.5b | image-text-to-text | — | not measured | not measured | galaxy-s26 cpu: fallback · 69.52 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-05galaxy-s26 gpu: pass · 86.45 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-08-25 | CARD.md · card.json | — |
m-lsd-tiny | image-segmentation | — | pass · 588.348 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | output_mismatch · 11.245 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | galaxy-s26 gpu_mldrift: pass · 6.174 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 · 2026-08-25galaxy-s26 npu_qnn: pass · 2.205 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · QAIRT (Hexagon, JIT on-device) · Android 16 · 2026-08-25raspberry-pi-5 cpu_xnnpack: pass · 106.619 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
mage-vl-cowork | image-text-to-text | — | not measured | not measured | iphone-18-pro cpu: pass · 10.5 tok/s decode stale — latest known 0.19.1iPhone 18 Pro (iPhone19,2) · litert-lm 0.19.0 · iOS 27.2 (24B5099f) · 2026-11-04iphone-18-pro gpu: pass stale — latest known 0.19.1iPhone 18 Pro (iPhone19,2) · litert-lm 0.19.0 · iOS 27.2 (24B5099f) · 2026-11-05iphone-18-pro gpu: pass stale — latest known 0.19.1iPhone 18 Pro (iPhone19,2) · litert-lm 0.19.0 · iOS 27.2 (24B5099f) · 2026-11-04iphone-18-pro gpu: pass · 35.26 tok/s decode stale — latest known 0.19.1iPhone 18 Pro (iPhone19,2) · litert-lm 0.19.0 · iOS 27.2 (24B5099f) · 2026-11-04mac-studio-m4-max cpu: pass stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.0 · macOS 27.0 (26A428) · 2026-11-04mac-studio-m4-max cpu: pass · 17.16 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.0 · macOS 27.0 (26A428) · 2026-11-04mac-studio-m4-max cpu: pass · 19.71 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.0 · macOS 27.0 (26A428) · 2026-11-04mac-studio-m4-max gpu: pass stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.0 · macOS 27.0 (26A428) · 2026-11-04mac-studio-m4-max gpu: pass stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.0 · macOS 27.0 (26A428) · 2026-11-04mac-studio-m4-max gpu: pass · 94.52 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.0 · macOS 27.0 (26A428) · 2026-11-04mac-studio-m4-max gpu: pass · 89.24 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.0 · macOS 27.0 (26A428) · 2026-11-04pixel-11-pro gpu: fallback · 7.2 tok/s decode stale — latest known 0.19.1Pixel 11 Pro · Tensor G6, Imagination PowerVR GPU · litert-lm 0.19.0 · Android 17 (CD1A.260905.001.B1) · 2026-11-04pixel-11-pro gpu: fallback stale — latest known 0.19.1Pixel 11 Pro · Tensor G6, Imagination PowerVR GPU · litert-lm 0.19.0 · Android 17 (CD1A.260905.001.B1) · 2026-11-05pixel-11-pro gpu: fallback stale — latest known 0.19.1Pixel 11 Pro · Tensor G6, Imagination PowerVR GPU · litert-lm 0.19.0 · Android 17 (CD1A.260905.001.B1) · 2026-11-04pixel-11-pro gpu: fallback · 7.81 tok/s decode stale — latest known 0.19.1Pixel 11 Pro · Tensor G6, Imagination PowerVR GPU · litert-lm 0.19.0 · Android 17 (CD1A.260905.001.B1) · 2026-11-04 | CARD.md · card.json | — |
mage-vl-cowork-vision-fp16 | image-text-to-text | — | not measured | not measured | galaxy-s26 cpu: fallback · 5.16 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU) · litert-lm 0.19.0 · Android 16 (BP4A.251205.006) · 2026-11-04galaxy-s26 gpu: fallback stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU) · litert-lm 0.19.0 · Android 16 (BP4A.251205.006) · 2026-11-04galaxy-s26 gpu: fallback stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU) · litert-lm 0.19.0 · Android 16 (BP4A.251205.006) · 2026-11-04galaxy-s26 gpu: fallback · 20.09 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU) · litert-lm 0.19.0 · Android 16 (BP4A.251205.006) · 2026-11-04galaxy-s26 gpu: fallback · 17.6 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU) · litert-lm 0.19.0 · Android 16 (BP4A.251205.006) · 2026-11-04mac-studio-m4-max gpu: pass stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.0 · macOS 27.0 (26A428) · 2026-11-04mac-studio-m4-max gpu: pass stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.0 · macOS 27.0 (26A428) · 2026-11-04 | CARD.md · card.json | — |
matcha-tts__dp_g2p_matcha_fp16 | text-to-speech | — | pass · 150.605 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | fallback · 136.315 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | raspberry-pi-5 cpu_xnnpack: pass · 21.997 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
matcha-tts__matcha_decoder_fp16 | text-to-speech | — | pass · 343.072 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | output_mismatch · 9.553 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | raspberry-pi-5 cpu_xnnpack: pass · 69.679 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
matcha-tts__matcha_textenc_fp16 | text-to-speech | — | pass · 229.993 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | pass · 5.992 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | raspberry-pi-5 cpu_xnnpack: pass · 38.679 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
matcha-tts__matcha_vocoder_fp16 | text-to-speech | — | pass · 21979.045 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | output_mismatch · 103.785 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | raspberry-pi-5 cpu_xnnpack: pass · 5824.518 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
mi-gan-512-places2 | image-to-image | — | pass · 3047.593 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | output_mismatch · 22.02 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | galaxy-s26 gpu_mldrift: pass · 26.645 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 · 2026-08-26galaxy-s26 npu_qnn: pass · 30.851 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · QAIRT (Hexagon, JIT on-device) · Android 16 · 2026-08-26raspberry-pi-5 cpu_xnnpack: pass · 1788.29 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
mimi__mimi_dec_tx_fp16 | neural-codec | — | pass · 40.158 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | output_mismatch · 4.683 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | galaxy-s26 gpu_mldrift: pass · 4.097 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 · 2026-08-26galaxy-s26 npu_qnn: pass · 2.088 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · QAIRT (Hexagon, JIT on-device) · Android 16 · 2026-08-26raspberry-pi-5 cpu_xnnpack: pass · 35.364 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
mimi__mimi_deconly_fp16 | neural-codec | — | pass · 1951.97 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | pass · 13.545 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | galaxy-s26 gpu_mldrift: pass · 23.902 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 · 2026-08-26galaxy-s26 npu_qnn: pass · 69.131 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · QAIRT (Hexagon, JIT on-device) · Android 16 · 2026-08-26raspberry-pi-5 cpu_xnnpack: pass · 586.3 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
mimi__mimi_enc_conv_fp16 | neural-codec | — | pass · 525.868 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | pass · 5.96 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | galaxy-s26 gpu_mldrift: pass · 14.688 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 · 2026-08-26galaxy-s26 npu_qnn: pass · 55.027 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · QAIRT (Hexagon, JIT on-device) · Android 16 · 2026-08-26raspberry-pi-5 cpu_xnnpack: pass · 298.623 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
mimi__mimi_enc_tx_fp16 | neural-codec | — | pass · 44.037 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | output_mismatch · 4.455 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | galaxy-s26 gpu_mldrift: pass · 4.35 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 · 2026-08-26galaxy-s26 npu_qnn: pass · 2.063 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · QAIRT (Hexagon, JIT on-device) · Android 16 · 2026-08-26raspberry-pi-5 cpu_xnnpack: pass · 36.564 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
minicpm5-1b | text-generation | — | not measured | not measured | mac-studio-m4-max cpu: pass stale — latest known 0.19.1Mac Studio (M4 Max) · litert-lm 0.17.0 · 2026-09-06 | CARD.md · card.json | — |
minicpm5-2b-int4 | text-generation | — | not measured | not measured | galaxy-s26 cpu: fallback · 15.56 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-08galaxy-s26 gpu: pass · 18.56 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-08iphone-17-pro cpu: pass stale — latest known 0.19.1iPhone 17 Pro · A19-Pro · litert-lm 0.15.0 · iOS 27.0 · 2026-09-08iphone-17-pro gpu: pass stale — latest known 0.19.1iPhone 17 Pro · A19-Pro · litert-lm 0.15.0 · iOS 27.0 · 2026-09-08mac-studio-m4-max cpu: pass · 31.12 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · litert-lm 0.17.0 · 2026-09-08mac-studio-m4-max gpu: pass · 92.79 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · litert-lm 0.17.0 · 2026-09-08 | CARD.md · card.json | — |
minicpm5-2b-int8 | text-generation | — | not measured | not measured | galaxy-s26 cpu: fallback · 11.67 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-08galaxy-s26 gpu: pass · 10.86 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-08mac-studio-m4-max cpu: pass · 30.02 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · litert-lm 0.17.0 · 2026-09-08mac-studio-m4-max gpu: pass · 74.65 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · litert-lm 0.17.0 · 2026-09-08 | CARD.md · card.json | — |
ministral3-3b | text-generation | — | not measured | not measured | galaxy-s26 cpu: fallback · 10.2 tok/s decodeGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm prebuilt-adac974b · Android 16 · 2026-10-05galaxy-s26 cpu: fallback · 8.48 tok/s decodeGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm prebuilt-adac974b · Android 16 · 2026-10-05galaxy-s26 gpu: pass · 22.09 tok/s decodeGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm prebuilt-adac974b · Android 16 · 2026-10-05galaxy-s26 gpu: pass · 11.26 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-08-24galaxy-s26 gpu: pass · 22.13 tok/s decodeGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm prebuilt-adac974b · Android 16 · 2026-10-05iphone-18-pro cpu: fallback · 15.52 tok/s decode stale — latest known 0.19.1iPhone 18 Pro (iPhone19,2) · litert-lm 0.17.0 · iOS 27.0 · 2026-10-05iphone-18-pro cpu: fallback · 18.08 tok/s decode stale — latest known 0.19.1iPhone 18 Pro (iPhone19,2) · litert-lm 0.17.0 · iOS 27.0 · 2026-10-05iphone-18-pro gpu: pass · 43.68 tok/s decode stale — latest known 0.19.1iPhone 18 Pro (iPhone19,2) · litert-lm 0.17.0 · iOS 27.0 · 2026-10-05iphone-18-pro gpu: pass · 43.99 tok/s decode stale — latest known 0.19.1iPhone 18 Pro (iPhone19,2) · litert-lm 0.17.0 · iOS 27.0 · 2026-10-05mac-studio-m4-max cpu: fallback · 17.77 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.17.1 · 2026-10-05mac-studio-m4-max cpu: fallback · 22.45 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.17.1 · 2026-10-05mac-studio-m4-max gpu: pass · 122.96 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.17.1 · 2026-10-05mac-studio-m4-max gpu: pass · 129.49 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.17.1 · 2026-10-05pixel-8a gpu: run_failedPixel 8a · Google Tensor G3 · litert-lm gallery-1.0.15 · 2026-07-23pixel-8a gpu: pass · 6.8 tok/s decode stale — latest known 0.19.1Pixel 8a · Tensor G3 · litert-lm 0.16.0 · 2026-08-17raspberry-pi-5 cpu: pass · 1.91 tok/s decode stale — latest known 0.19.1Raspberry Pi 5 Model B Rev 1.1 · Broadcom BCM2712 · litert-lm 0.16.1 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-09-01 | CARD.md · card.json | — |
ministral3-3b-reasoning | text-generation | — | not measured | not measured | mac-studio-m4-max gpu: pass · 94.65 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.16.0 · 2026-08-12raspberry-pi-5 cpu: pass · 1.9 tok/s decode stale — latest known 0.19.1Raspberry Pi 5 Model B Rev 1.1 · Broadcom BCM2712 · litert-lm 0.16.1 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-09-01 | CARD.md · card.json | — |
ministral3-8b-int4 | text-generation | — | not measured | not measured | galaxy-s26 gpu: pass · 5.6425 tok/s decodeGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm prebuilt-adac974b · Android 16 (BP4A.251205.006) · 2026-10-15mac-studio-m4-max cpu: pass · 12.93 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.18.0 · macOS 27.0 (26A428) · 2026-10-15mac-studio-m4-max gpu: pass stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.17.1 · macOS 27.0 (26A428) · 2026-10-14mac-studio-m4-max gpu: pass · 61.74 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.18.0 · macOS 27.0 (26A428) · 2026-10-15pixel-8a gpu: load_failedPixel 8a · Tensor G3 · litert-lm prebuilt-adac974b · Android 16 (CP1A.260505.005) · 2026-10-14 | CARD.md · card.json | — |
ministral3-8b-int8 | text-generation | — | not measured | not measured | mac-studio-m4-max cpu: pass · 11.49 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.18.0 · macOS 27.0 (26A428) · 2026-10-15mac-studio-m4-max gpu: pass stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.17.1 · macOS 27.0 (26A428) · 2026-10-14mac-studio-m4-max gpu: pass · 46.66 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.18.0 · macOS 27.0 (26A428) · 2026-10-15 | CARD.md · card.json | — |
mnn-pocket-0.3b-dec__mnn-pocket-0.3b-dec_l1024_fp16w | text-classification | — | not measured | not measured | galaxy-s26 cpu_xnnpack: fallbackGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-07galaxy-s26 gpu_mldrift: passGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-07galaxy-s26 gpu_mldrift: passGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-07galaxy-s26 gpu_mldrift: pass · 371.5675 msGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-08galaxy-s26 gpu_mldrift: pass · 530.277 msGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-08mac-studio-m4-max cpu_xnnpack: fallbackMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert 2.3.0 · macOS 27.0 (26A428) · 2026-11-07mac-studio-m4-max cpu_xnnpack: fallback · 380.9292295045452 msMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert 2.3.0 · macOS 27.0 (26A428) · 2026-11-07mac-studio-m4-max gpu_metal_mac: passMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert 2.3.0 · macOS 27.0 (26A428) · 2026-11-07mac-studio-m4-max gpu_metal_mac: pass · 90.18472899333574 msMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert 2.3.0 · macOS 27.0 (26A428) · 2026-11-07pixel-11-pro cpu_xnnpack: fallbackPixel 11 Pro · Google Tensor G6 · litert 2.3.0 · Android 17 (CD1A.260905.001.B1) · 2026-11-07pixel-11-pro cpu_xnnpack: fallback · 2871.7195 msPixel 11 Pro · Google Tensor G6 · litert 2.3.0 · Android 17 (CD1A.260905.001.B1) · 2026-11-08pixel-11-pro gpu_mldrift: passPixel 11 Pro · Google Tensor G6 · litert 2.3.0 · Android 17 (CD1A.260905.001.B1) · 2026-11-07pixel-11-pro gpu_mldrift: pass · 1020.7149999999999 msPixel 11 Pro · Google Tensor G6 · litert 2.3.0 · Android 17 (CD1A.260905.001.B1) · 2026-11-08 | CARD.md · card.json | — |
mnn-pocket-0.3b-dec__mnn-pocket-0.3b-dec_l1024_fp16w_npu | text-classification | — | not measured | not measured | galaxy-s26 npu_qnn: passGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · QAIRT 2.50.0.260828221209 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-07galaxy-s26 npu_qnn: pass · 206.6945 msGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · QAIRT 2.50.0.260828221209 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-07galaxy-s26 npu_qnn: pass · 205.5455 msGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · QAIRT 2.50.0.260828221209 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-07galaxy-s26 npu_qnn: pass · 206.0425 msGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · QAIRT 2.50.0.260828221209 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-07mac-studio-m4-max cpu_xnnpack: fallbackMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert 2.3.0 · macOS 27.0 (26A428) · 2026-11-07mac-studio-m4-max gpu_metal_mac: passMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert 2.3.0 · macOS 27.0 (26A428) · 2026-11-07 | CARD.md · card.json | — |
mnn-pocket-0.3b-dec__mnn-pocket-0.3b-dec_l1024_fp16w_npu_sm8850 | text-classification | — | not measured | not measured | galaxy-s26 npu_qnn: passGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · QAIRT 2.50.0.260828221209 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-07galaxy-s26 npu_qnn: pass · 203.971 msGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · QAIRT 2.50.0.260828221209 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-07galaxy-s26 npu_qnn: pass · 203.44 msGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · QAIRT 2.50.0.260828221209 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-07galaxy-s26 npu_qnn: pass · 203.41 msGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · QAIRT 2.50.0.260828221209 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-07 | CARD.md · card.json | — |
mnn-pocket-0.3b-dec__mnn-pocket-0.3b-dec_l128_fp16w | text-classification | — | not measured | not measured | galaxy-s26 cpu_xnnpack: fallbackGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-07galaxy-s26 gpu_mldrift: passGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-07galaxy-s26 gpu_mldrift: passGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-07mac-studio-m4-max cpu_xnnpack: fallbackMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert 2.3.0 · macOS 27.0 (26A428) · 2026-11-07mac-studio-m4-max gpu_metal_mac: passMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert 2.3.0 · macOS 27.0 (26A428) · 2026-11-07pixel-11-pro cpu_xnnpack: fallbackPixel 11 Pro · Google Tensor G6 · litert 2.3.0 · Android 17 (CD1A.260905.001.B1) · 2026-11-08pixel-11-pro gpu_mldrift: passPixel 11 Pro · Google Tensor G6 · litert 2.3.0 · Android 17 (CD1A.260905.001.B1) · 2026-11-07 | CARD.md · card.json | — |
mnn-pocket-0.3b-dec__mnn-pocket-0.3b-dec_l128_fp16w_npu | text-classification | — | not measured | not measured | galaxy-s26 npu_qnn: passGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · QAIRT 2.50.0.260828221209 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-07galaxy-s26 npu_qnn: pass · 20.689999999999998 msGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · QAIRT 2.50.0.260828221209 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-07galaxy-s26 npu_qnn: pass · 20.9 msGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · QAIRT 2.50.0.260828221209 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-07mac-studio-m4-max cpu_xnnpack: fallbackMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert 2.3.0 · macOS 27.0 (26A428) · 2026-11-07mac-studio-m4-max gpu_metal_mac: passMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert 2.3.0 · macOS 27.0 (26A428) · 2026-11-07 | CARD.md · card.json | — |
mnn-pocket-0.3b-dec__mnn-pocket-0.3b-dec_l128_fp16w_npu_sm8850 | text-classification | — | not measured | not measured | galaxy-s26 npu_qnn: passGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · QAIRT 2.50.0.260828221209 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-07galaxy-s26 npu_qnn: pass · 20.5775 msGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · QAIRT 2.50.0.260828221209 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-07galaxy-s26 npu_qnn: pass · 20.545 msGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · QAIRT 2.50.0.260828221209 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-07 | CARD.md · card.json | — |
mnn-pocket-0.3b-dec__mnn-pocket-0.3b-dec_l128_fp16w_tensor | text-classification | — | not measured | not measured | mac-studio-m4-max cpu_xnnpack: fallbackMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert 2.3.0 · macOS 27.0 (26A428) · 2026-11-08mac-studio-m4-max gpu_metal_mac: passMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert 2.3.0 · macOS 27.0 (26A428) · 2026-11-08pixel-11-pro npu_google_tensor: passPixel 11 Pro · Google Tensor G6 · litert 2.3.0 · LiteRT Google Tensor NPU runtime 2.3.0 (Maven litert-npu-runtime-google-tensor:2.3.0; compiler plugin 6b1432e7…, dispatch e3814400…) · Android 17 (CD1A.260905.001.B1) · 2026-11-08 | CARD.md · card.json | — |
mnn-pocket-0.3b-dec__mnn-pocket-0.3b-dec_l2048_fp16w | text-classification | — | not measured | not measured | galaxy-s26 cpu_xnnpack: fallbackGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-07galaxy-s26 gpu_mldrift: passGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-07galaxy-s26 gpu_mldrift: passGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-07mac-studio-m4-max cpu_xnnpack: fallbackMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert 2.3.0 · macOS 27.0 (26A428) · 2026-11-07mac-studio-m4-max gpu_metal_mac: passMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert 2.3.0 · macOS 27.0 (26A428) · 2026-11-07pixel-11-pro cpu_xnnpack: fallbackPixel 11 Pro · Google Tensor G6 · litert 2.3.0 · Android 17 (CD1A.260905.001.B1) · 2026-11-07pixel-11-pro gpu_mldrift: passPixel 11 Pro · Google Tensor G6 · litert 2.3.0 · Android 17 (CD1A.260905.001.B1) · 2026-11-07 | CARD.md · card.json | — |
mnn-pocket-0.3b-dec__mnn-pocket-0.3b-dec_l2048_fp16w_npu | text-classification | — | not measured | not measured | galaxy-s26 npu_qnn: run_failedGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · QAIRT 2.50.0.260828221209 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-07mac-studio-m4-max cpu_xnnpack: fallbackMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert 2.3.0 · macOS 27.0 (26A428) · 2026-11-07mac-studio-m4-max gpu_metal_mac: passMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert 2.3.0 · macOS 27.0 (26A428) · 2026-11-07 | CARD.md · card.json | — |
mnn-pocket-0.3b-dec__mnn-pocket-0.3b-dec_l2048_fp16w_npu_sm8850 | text-classification | — | not measured | not measured | galaxy-s26 npu_qnn: passGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · QAIRT 2.50.0.260828221209 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-07galaxy-s26 npu_qnn: pass · 882.7855 msGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · QAIRT 2.50.0.260828221209 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-07galaxy-s26 npu_qnn: pass · 878.2395 msGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · QAIRT 2.50.0.260828221209 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-07galaxy-s26 npu_qnn: pass · 876.785 msGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · QAIRT 2.50.0.260828221209 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-07 | CARD.md · card.json | — |
mnn-pocket-0.3b-dec__mnn-pocket-0.3b-dec_l256_fp16w | text-classification | — | not measured | not measured | galaxy-s26 cpu_xnnpack: fallbackGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-07galaxy-s26 gpu_mldrift: passGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-07galaxy-s26 gpu_mldrift: passGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-07mac-studio-m4-max cpu_xnnpack: fallbackMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert 2.3.0 · macOS 27.0 (26A428) · 2026-11-07mac-studio-m4-max gpu_metal_mac: passMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert 2.3.0 · macOS 27.0 (26A428) · 2026-11-07pixel-11-pro cpu_xnnpack: fallbackPixel 11 Pro · Google Tensor G6 · litert 2.3.0 · Android 17 (CD1A.260905.001.B1) · 2026-11-07pixel-11-pro gpu_mldrift: passPixel 11 Pro · Google Tensor G6 · litert 2.3.0 · Android 17 (CD1A.260905.001.B1) · 2026-11-07 | CARD.md · card.json | — |
mnn-pocket-0.3b-dec__mnn-pocket-0.3b-dec_l256_fp16w_npu | text-classification | — | not measured | not measured | galaxy-s26 npu_qnn: passGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · QAIRT 2.50.0.260828221209 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-07galaxy-s26 npu_qnn: pass · 39.221500000000006 msGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · QAIRT 2.50.0.260828221209 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-07galaxy-s26 npu_qnn: pass · 39.320499999999996 msGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · QAIRT 2.50.0.260828221209 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-07mac-studio-m4-max cpu_xnnpack: fallbackMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert 2.3.0 · macOS 27.0 (26A428) · 2026-11-07mac-studio-m4-max gpu_metal_mac: passMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert 2.3.0 · macOS 27.0 (26A428) · 2026-11-07 | CARD.md · card.json | — |
mnn-pocket-0.3b-dec__mnn-pocket-0.3b-dec_l256_fp16w_npu_sm8850 | text-classification | — | not measured | not measured | galaxy-s26 npu_qnn: passGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · QAIRT 2.50.0.260828221209 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-07galaxy-s26 npu_qnn: pass · 39.4925 msGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · QAIRT 2.50.0.260828221209 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-07galaxy-s26 npu_qnn: pass · 39.579499999999996 msGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · QAIRT 2.50.0.260828221209 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-07 | CARD.md · card.json | — |
mnn-pocket-0.3b-dec__mnn-pocket-0.3b-dec_l256_fp16w_tensor | text-classification | — | not measured | not measured | mac-studio-m4-max cpu_xnnpack: fallbackMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert 2.3.0 · macOS 27.0 (26A428) · 2026-11-08mac-studio-m4-max gpu_metal_mac: passMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert 2.3.0 · macOS 27.0 (26A428) · 2026-11-08pixel-11-pro npu_google_tensor: passPixel 11 Pro · Google Tensor G6 · litert 2.3.0 · LiteRT Google Tensor NPU runtime 2.3.0 (Maven litert-npu-runtime-google-tensor:2.3.0; compiler plugin 6b1432e7…, dispatch e3814400…) · Android 17 (CD1A.260905.001.B1) · 2026-11-08 | CARD.md · card.json | — |
mnn-pocket-0.3b-dec__mnn-pocket-0.3b-dec_l512_fp16w | text-classification | — | not measured | not measured | galaxy-s26 cpu_xnnpack: fallbackGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-07galaxy-s26 gpu_mldrift: passGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-07galaxy-s26 gpu_mldrift: passGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-07mac-studio-m4-max cpu_xnnpack: fallbackMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert 2.3.0 · macOS 27.0 (26A428) · 2026-11-07mac-studio-m4-max gpu_metal_mac: passMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert 2.3.0 · macOS 27.0 (26A428) · 2026-11-07pixel-11-pro cpu_xnnpack: fallbackPixel 11 Pro · Google Tensor G6 · litert 2.3.0 · Android 17 (CD1A.260905.001.B1) · 2026-11-07pixel-11-pro gpu_mldrift: passPixel 11 Pro · Google Tensor G6 · litert 2.3.0 · Android 17 (CD1A.260905.001.B1) · 2026-11-07 | CARD.md · card.json | — |
mnn-pocket-0.3b-dec__mnn-pocket-0.3b-dec_l512_fp16w_npu | text-classification | — | not measured | not measured | galaxy-s26 npu_qnn: passGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · QAIRT 2.50.0.260828221209 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-07galaxy-s26 npu_qnn: pass · 109.8875 msGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · QAIRT 2.50.0.260828221209 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-07galaxy-s26 npu_qnn: pass · 113.0645 msGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · QAIRT 2.50.0.260828221209 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-07mac-studio-m4-max cpu_xnnpack: fallbackMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert 2.3.0 · macOS 27.0 (26A428) · 2026-11-07mac-studio-m4-max gpu_metal_mac: passMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert 2.3.0 · macOS 27.0 (26A428) · 2026-11-07 | CARD.md · card.json | — |
mnn-pocket-0.3b-dec__mnn-pocket-0.3b-dec_l512_fp16w_npu_sm8850 | text-classification | — | not measured | not measured | galaxy-s26 npu_qnn: passGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · QAIRT 2.50.0.260828221209 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-07galaxy-s26 npu_qnn: pass · 98.275 msGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · QAIRT 2.50.0.260828221209 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-07galaxy-s26 npu_qnn: pass · 100.3455 msGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · QAIRT 2.50.0.260828221209 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-07 | CARD.md · card.json | — |
mnn-pocket-0.3b-dec__mnn-pocket-0.3b-dec_l64_fp16w | text-classification | — | load_failed stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-11-08 | load_failed stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-11-08 | galaxy-s26 cpu_xnnpack: fallbackGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-07galaxy-s26 cpu_xnnpack: fallback · 41.0525 msGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-07galaxy-s26 gpu_mldrift: passGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-07galaxy-s26 gpu_mldrift: passGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-07galaxy-s26 gpu_mldrift: pass · 28.186 msGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-07galaxy-s26 gpu_mldrift: pass · 37.593 msGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-07mac-studio-m4-max cpu_xnnpack: fallbackMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert 2.3.0 · macOS 27.0 (26A428) · 2026-11-07mac-studio-m4-max cpu_xnnpack: fallback · 29.71714601153508 msMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert 2.3.0 · macOS 27.0 (26A428) · 2026-11-08mac-studio-m4-max gpu_metal_mac: passMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert 2.3.0 · macOS 27.0 (26A428) · 2026-11-07mac-studio-m4-max gpu_metal_mac: pass · 12.622374997590668 msMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert 2.3.0 · macOS 27.0 (26A428) · 2026-11-07pixel-11-pro cpu_xnnpack: fallbackPixel 11 Pro · Google Tensor G6 · litert 2.3.0 · Android 17 (CD1A.260905.001.B1) · 2026-11-08pixel-11-pro cpu_xnnpack: fallback · 128.7415 msPixel 11 Pro · Google Tensor G6 · litert 2.3.0 · Android 17 (CD1A.260905.001.B1) · 2026-11-08pixel-11-pro gpu_mldrift: passPixel 11 Pro · Google Tensor G6 · litert 2.3.0 · Android 17 (CD1A.260905.001.B1) · 2026-11-07pixel-11-pro gpu_mldrift: pass · 120.898 msPixel 11 Pro · Google Tensor G6 · litert 2.3.0 · Android 17 (CD1A.260905.001.B1) · 2026-11-07 | CARD.md · card.json | — |
mnn-pocket-0.3b-dec__mnn-pocket-0.3b-dec_l64_fp16w_npu | text-classification | — | not measured | not measured | galaxy-s26 npu_qnn: passGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · QAIRT 2.50.0.260828221209 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-07galaxy-s26 npu_qnn: pass · 11.399000000000001 msGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · QAIRT 2.50.0.260828221209 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-07mac-studio-m4-max cpu_xnnpack: fallbackMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert 2.3.0 · macOS 27.0 (26A428) · 2026-11-07mac-studio-m4-max gpu_metal_mac: passMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert 2.3.0 · macOS 27.0 (26A428) · 2026-11-07 | CARD.md · card.json | — |
mnn-pocket-0.3b-dec__mnn-pocket-0.3b-dec_l64_fp16w_npu_sm8850 | text-classification | — | not measured | not measured | galaxy-s26 npu_qnn: passGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · QAIRT 2.50.0.260828221209 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-07galaxy-s26 npu_qnn: pass · 11.2805 msGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · QAIRT 2.50.0.260828221209 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-07 | CARD.md · card.json | — |
mnn-pocket-0.3b-dec__mnn-pocket-0.3b-dec_l64_fp16w_tensor | text-classification | — | not measured | not measured | mac-studio-m4-max cpu_xnnpack: fallbackMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert 2.3.0 · macOS 27.0 (26A428) · 2026-11-08mac-studio-m4-max gpu_metal_mac: passMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert 2.3.0 · macOS 27.0 (26A428) · 2026-11-08pixel-11-pro npu_google_tensor: passPixel 11 Pro · Google Tensor G6 · litert 2.3.0 · LiteRT Google Tensor NPU runtime 2.3.0 (Maven litert-npu-runtime-google-tensor:2.3.0; compiler plugin 6b1432e7…, dispatch e3814400…) · Android 17 (CD1A.260905.001.B1) · 2026-11-08 | CARD.md · card.json | — |
mnn-pocket-0.3b-dec__mnn-pocket-0.3b-dec_l768_fp16w | text-classification | — | not measured | not measured | galaxy-s26 cpu_xnnpack: fallbackGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-08galaxy-s26 cpu_xnnpack: fallback · 673.48 msGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-08galaxy-s26 gpu_mldrift: passGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-08galaxy-s26 gpu_mldrift: passGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-08galaxy-s26 gpu_mldrift: pass · 247.837 msGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-08galaxy-s26 gpu_mldrift: pass · 352.4285 msGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-08mac-studio-m4-max cpu_xnnpack: fallbackMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert 2.3.0 · macOS 27.0 (26A428) · 2026-11-08mac-studio-m4-max cpu_xnnpack: fallback · 221.45497899327893 msMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert 2.3.0 · macOS 27.0 (26A428) · 2026-11-08mac-studio-m4-max gpu_metal_mac: passMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert 2.3.0 · macOS 27.0 (26A428) · 2026-11-08mac-studio-m4-max gpu_metal_mac: pass · 59.80418750550598 msMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert 2.3.0 · macOS 27.0 (26A428) · 2026-11-08pixel-11-pro cpu_xnnpack: fallbackPixel 11 Pro · Google Tensor G6 · litert 2.3.0 · Android 17 (CD1A.260905.001.B1) · 2026-11-08pixel-11-pro cpu_xnnpack: fallback · 1900.8760000000002 msPixel 11 Pro · Google Tensor G6 · litert 2.3.0 · Android 17 (CD1A.260905.001.B1) · 2026-11-08pixel-11-pro gpu_mldrift: passPixel 11 Pro · Google Tensor G6 · litert 2.3.0 · Android 17 (CD1A.260905.001.B1) · 2026-11-08pixel-11-pro gpu_mldrift: pass · 716.4125 msPixel 11 Pro · Google Tensor G6 · litert 2.3.0 · Android 17 (CD1A.260905.001.B1) · 2026-11-08 | CARD.md · card.json | — |
mnn-pocket-0.3b-dec__mnn-pocket-0.3b-dec_l768_fp16w_npu | text-classification | — | not measured | not measured | galaxy-s26 npu_qnn: passGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · QAIRT 2.50.0.260828221209 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-08galaxy-s26 npu_qnn: pass · 142.4675 msGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · QAIRT 2.50.0.260828221209 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-08mac-studio-m4-max cpu_xnnpack: fallbackMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert 2.3.0 · macOS 27.0 (26A428) · 2026-11-08mac-studio-m4-max gpu_metal_mac: passMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert 2.3.0 · macOS 27.0 (26A428) · 2026-11-08 | CARD.md · card.json | — |
mnn-pocket-0.3b-dec__mnn-pocket-0.3b-dec_l768_fp16w_npu_sm8850 | text-classification | — | not measured | not measured | galaxy-s26 npu_qnn: passGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · QAIRT 2.50.0.260828221209 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-08galaxy-s26 npu_qnn: pass · 140.067 msGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · QAIRT 2.50.0.260828221209 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-08 | CARD.md · card.json | — |
mnn-pocket-0.3b-dec__mnn-pocket-0.3b-dec_l96_fp16w | text-classification | — | not measured | not measured | galaxy-s26 cpu_xnnpack: fallbackGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-07galaxy-s26 cpu_xnnpack: fallback · 60.884 msGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-07galaxy-s26 gpu_mldrift: passGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-07galaxy-s26 gpu_mldrift: passGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-07galaxy-s26 gpu_mldrift: pass · 43.6315 msGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-07galaxy-s26 gpu_mldrift: pass · 48.248000000000005 msGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-07mac-studio-m4-max cpu_xnnpack: fallbackMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert 2.3.0 · macOS 27.0 (26A428) · 2026-11-07mac-studio-m4-max cpu_xnnpack: fallback · 41.12781251023989 msMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert 2.3.0 · macOS 27.0 (26A428) · 2026-11-07mac-studio-m4-max gpu_metal_mac: passMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert 2.3.0 · macOS 27.0 (26A428) · 2026-11-07mac-studio-m4-max gpu_metal_mac: pass · 13.381667013163678 msMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert 2.3.0 · macOS 27.0 (26A428) · 2026-11-07pixel-11-pro cpu_xnnpack: fallbackPixel 11 Pro · Google Tensor G6 · litert 2.3.0 · Android 17 (CD1A.260905.001.B1) · 2026-11-08pixel-11-pro cpu_xnnpack: fallback · 179.4755 msPixel 11 Pro · Google Tensor G6 · litert 2.3.0 · Android 17 (CD1A.260905.001.B1) · 2026-11-08pixel-11-pro gpu_mldrift: passPixel 11 Pro · Google Tensor G6 · litert 2.3.0 · Android 17 (CD1A.260905.001.B1) · 2026-11-07pixel-11-pro gpu_mldrift: pass · 129.0575 msPixel 11 Pro · Google Tensor G6 · litert 2.3.0 · Android 17 (CD1A.260905.001.B1) · 2026-11-07 | CARD.md · card.json | — |
mnn-pocket-0.3b-dec__mnn-pocket-0.3b-dec_l96_fp16w_npu | text-classification | — | not measured | not measured | galaxy-s26 npu_qnn: passGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · QAIRT 2.50.0.260828221209 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-07galaxy-s26 npu_qnn: pass · 15.92 msGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · QAIRT 2.50.0.260828221209 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-07galaxy-s26 npu_qnn: pass · 15.763 msGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · QAIRT 2.50.0.260828221209 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-07mac-studio-m4-max cpu_xnnpack: fallbackMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert 2.3.0 · macOS 27.0 (26A428) · 2026-11-07mac-studio-m4-max gpu_metal_mac: passMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert 2.3.0 · macOS 27.0 (26A428) · 2026-11-07 | CARD.md · card.json | — |
mnn-pocket-0.3b-dec__mnn-pocket-0.3b-dec_l96_fp16w_npu_sm8850 | text-classification | — | not measured | not measured | galaxy-s26 npu_qnn: passGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · QAIRT 2.50.0.260828221209 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-07galaxy-s26 npu_qnn: pass · 15.956 msGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · QAIRT 2.50.0.260828221209 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-07galaxy-s26 npu_qnn: pass · 16.2065 msGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.3.0 · QAIRT 2.50.0.260828221209 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-07 | CARD.md · card.json | — |
mnn-pocket-0.3b-dec__mnn-pocket-0.3b-dec_l96_fp16w_tensor | text-classification | — | not measured | not measured | mac-studio-m4-max cpu_xnnpack: fallbackMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert 2.3.0 · macOS 27.0 (26A428) · 2026-11-08mac-studio-m4-max gpu_metal_mac: passMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert 2.3.0 · macOS 27.0 (26A428) · 2026-11-08pixel-11-pro npu_google_tensor: passPixel 11 Pro · Google Tensor G6 · litert 2.3.0 · LiteRT Google Tensor NPU runtime 2.3.0 (Maven litert-npu-runtime-google-tensor:2.3.0; compiler plugin 6b1432e7…, dispatch e3814400…) · Android 17 (CD1A.260905.001.B1) · 2026-11-08 | CARD.md · card.json | — |
mnn-pocket-0.3b-int8 | text-generation | — | not measured | not measured | galaxy-s26 cpu: fallbackGalaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU) · litert-lm 0.19.1 · Android 16 (BP4A.251205.006) · 2026-11-07galaxy-s26 cpu: fallback · 53.242 tok/s decodeGalaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU) · litert-lm 0.19.1 · Android 16 (BP4A.251205.006) · 2026-11-06galaxy-s26 gpu: pass · 105.5755 tok/s decodeGalaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU) · litert-lm 0.19.1 · Android 16 (BP4A.251205.006) · 2026-11-06mac-studio-m4-max cpu: passMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.1 · macOS 27.0 (26A428) · 2026-11-07mac-studio-m4-max cpu: pass · 95.85 tok/s decodeMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.1 · macOS 27.0 (26A428) · 2026-11-06mac-studio-m4-max gpu: passMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.1 · macOS 27.0 (26A428) · 2026-11-07mac-studio-m4-max gpu: pass · 490.505 tok/s decodeMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.1 · macOS 27.0 (26A428) · 2026-11-06pixel-11-pro cpu: fallbackPixel 11 Pro · Tensor G6, Imagination PowerVR GPU · litert-lm 0.19.1 · Android 17 (CD1A.260905.001.B1) · 2026-11-07pixel-11-pro cpu: fallback · 45.418 tok/s decodePixel 11 Pro · Tensor G6, Imagination PowerVR GPU · litert-lm 0.19.1 · Android 17 (CD1A.260905.001.B1) · 2026-11-07pixel-11-pro gpu: pass · 28.272 tok/s decodePixel 11 Pro · Tensor G6, Imagination PowerVR GPU · litert-lm 0.19.1 · Android 17 (CD1A.260905.001.B1) · 2026-11-07 | CARD.md · card.json | — |
mnn-pocket-0.3b-int8-ekv2048 | text-generation | — | not measured | not measured | galaxy-s26 cpu: fallbackGalaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU) · litert-lm 0.19.1 · Android 16 (BP4A.251205.006) · 2026-11-07galaxy-s26 cpu: fallback · 87.405 tok/s decodeGalaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU) · litert-lm 0.19.1 · Android 16 (BP4A.251205.006) · 2026-11-06galaxy-s26 gpu: passGalaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU) · litert-lm 0.19.1 · Android 16 (BP4A.251205.006) · 2026-11-07galaxy-s26 gpu: pass · 120.0495 tok/s decodeGalaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU) · litert-lm 0.19.1 · Android 16 (BP4A.251205.006) · 2026-11-06mac-studio-m4-max cpu: passMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.1 · macOS 27.0 (26A428) · 2026-11-06mac-studio-m4-max cpu: pass · 116.025 tok/s decodeMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.1 · macOS 27.0 (26A428) · 2026-11-06mac-studio-m4-max gpu: passMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.1 · macOS 27.0 (26A428) · 2026-11-06mac-studio-m4-max gpu: pass · 490.94 tok/s decodeMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.1 · macOS 27.0 (26A428) · 2026-11-06pixel-11-pro cpu: fallbackPixel 11 Pro · Tensor G6, Imagination PowerVR GPU · litert-lm 0.19.1 · Android 17 (CD1A.260905.001.B1) · 2026-11-07pixel-11-pro cpu: fallback · 76.3625 tok/s decodePixel 11 Pro · Tensor G6, Imagination PowerVR GPU · litert-lm 0.19.1 · Android 17 (CD1A.260905.001.B1) · 2026-11-07pixel-11-pro gpu: passPixel 11 Pro · Tensor G6, Imagination PowerVR GPU · litert-lm 0.19.1 · Android 17 (CD1A.260905.001.B1) · 2026-11-07pixel-11-pro gpu: pass · 33.283 tok/s decodePixel 11 Pro · Tensor G6, Imagination PowerVR GPU · litert-lm 0.19.1 · Android 17 (CD1A.260905.001.B1) · 2026-11-07 | CARD.md · card.json | — |
mnn-pocket-0.3b-qualcomm-npu-ekv2048 | text-generation | — | not measured | not measured | galaxy-s26 npu: fallbackGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.19.1 · QAIRT 2.47.0.260601114230 (QNN JIT libraries) · Android 16 (BP4A.251205.006) · 2026-11-07galaxy-s26 npu: fallbackGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.19.1 · QAIRT 2.47.0.260601114230 (QNN JIT libraries) · Android 16 (BP4A.251205.006) · 2026-11-07galaxy-s26 npu: fallback · 147.2525 tok/s decodeGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.19.1 · QAIRT 2.47.0.260601114230 (QNN JIT libraries) · Android 16 (BP4A.251205.006) · 2026-11-07 | CARD.md · card.json | — |
mnn-pocket-0.3b-qualcomm-sm8850-ekv2048 | text-generation | — | not measured | not measured | galaxy-s26 npu: fallbackGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.19.1 · QAIRT 2.47.0.260601114230 (QNN libraries on the phone; no compiler plugin) · Android 16 (BP4A.251205.006) · 2026-11-07galaxy-s26 npu: fallback · 157.7375 tok/s decodeGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.19.1 · QAIRT 2.47.0.260601114230 (QNN libraries on the phone; no compiler plugin) · Android 16 (BP4A.251205.006) · 2026-11-07 | CARD.md · card.json | — |
modernbert-ja-310m-decision__modernbert_ja_310m_decision_s256_wfp16 | typed decisions in Japanese (question + state + candidates -> one logit per (context, candidate) pair; softmax over a question's candidates = choice / score / true-false; text-classification) | gpu_mldrift @ litert 2.2.0: 91.9% · 155 partition(s) | run_failed stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-10-02 | load_failed stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-10-02 | galaxy-s26 gpu_mldrift: pass · 229.156823 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (SDK 36) · 2026-10-02galaxy-s26 npu_qnn: output_mismatch · 41.726354 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (SDK 36) · 2026-10-02galaxy-s26 npu_qnn: output_mismatch · 44.617136 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (SDK 36) · 2026-10-02 | CARD.md · card.json | — |
modernbert-ja-310m-decision__modernbert_ja_310m_decision_s512_wfp16 | typed decisions in Japanese (question + state + candidates -> one logit per (context, candidate) pair; softmax over a question's candidates = choice / score / true-false; text-classification) | gpu_mldrift @ litert 2.2.0: 91.9% · 155 partition(s) | run_failed stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-10-02 | load_failed stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-10-02 | galaxy-s26 gpu_mldrift: output_mismatch · 230.960469 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (SDK 36) · 2026-10-02galaxy-s26 gpu_mldrift: output_mismatch · 252.711458 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (SDK 36) · 2026-10-02galaxy-s26 gpu_mldrift: output_mismatch · 402.7025 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (SDK 36) · 2026-10-02galaxy-s26 gpu_mldrift: pass · 416.436562 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (SDK 36) · 2026-10-02galaxy-s26 npu_qnn: output_mismatch · 142.595885 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (SDK 36) · 2026-10-02 | CARD.md · card.json | — |
modnet | image-segmentation | — | pass · 480.658 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | pass · 20.715 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | raspberry-pi-5 cpu_xnnpack: pass · 602.838 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
mordant-3b-think-int8 | text-generation | — | not measured | not measured | galaxy-s26 cpu: fallback · 7.7 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-05galaxy-s26 gpu: pass · 9.59 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-05mac-studio-m4-max cpu: pass · 20.18 tok/s decode stale — latest known 0.19.1mac-studio-m4-max · Apple M4 Max · litert-lm 0.16.0 · 2026-08-25mac-studio-m4-max gpu: pass · 71.85 tok/s decode stale — latest known 0.19.1mac-studio-m4-max · Apple M4 Max · litert-lm 0.16.0 · 2026-08-25 | CARD.md · card.json | — |
movinet-a0-stream | video-classification | — | pass · 7.13 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | output_mismatch · 13.255 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | raspberry-pi-5 cpu_xnnpack: pass · 4.628 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
nafnet-gopro-width32 | image-to-image | — | pass · 2712.535 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | pass · 20.842 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | galaxy-s26 gpu_mldrift: pass · 30.894 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 · 2026-08-26galaxy-s26 npu_qnn: pass · 18.583 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · QAIRT (Hexagon, JIT on-device) · Android 16 · 2026-08-26raspberry-pi-5 cpu_xnnpack: pass · 820.934 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
nafnet-sidd-width32 | image-to-image | — | pass · 2825.957 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | output_mismatch · 23.618 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | galaxy-s26 gpu_mldrift: pass · 37.258 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 · 2026-08-26galaxy-s26 npu_qnn: pass · 26.295 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · QAIRT (Hexagon, JIT on-device) · Android 16 · 2026-08-26raspberry-pi-5 cpu_xnnpack: pass · 1039.801 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
nanbeige4.1-3b | text-generation | — | not measured | not measured | galaxy-s26 cpu: fallback · 10.58 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-05mac-studio-m4-max gpu: pass stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.14.0 · 2026-07-22raspberry-pi-5 cpu: pass · 2.49 tok/s decode stale — latest known 0.19.1Raspberry Pi 5 Model B Rev 1.1 · Broadcom BCM2712 · litert-lm 0.16.1 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-09-02 | CARD.md · card.json | — |
nanbeige4.2-3b | text-generation | — | not measured | not measured | galaxy-s26 cpu: fallback · 4.1 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-05galaxy-s26 gpu: load_failed stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-08-24mac-studio-m4-max cpu: pass stale — latest known 0.19.1Mac Studio (M4 Max) · litert-lm 0.17.0 · 2026-09-06raspberry-pi-5 cpu: pass · 1.06 tok/s decode stale — latest known 0.19.1Raspberry Pi 5 Model B Rev 1.1 · Broadcom BCM2712 · litert-lm 0.16.1 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-09-01 | CARD.md · card.json | — |
nanbeige4.2-3b-fp32act | text-generation | — | not measured | not measured | galaxy-s26 cpu: pass · 3.45 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-08-27galaxy-s26 gpu: pass · 3.31 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-08-27iphone-17-pro cpu: pass stale — latest known 0.19.1iPhone 17 Pro · A19-Pro · litert-lm 0.15.0 · iOS 27.0 · 2026-08-28iphone-17-pro gpu: pass stale — latest known 0.19.1iPhone 17 Pro · A19-Pro · litert-lm 0.15.0 · iOS 27.0 · 2026-08-28raspberry-pi-5 cpu: pass · 1.07 tok/s decode stale — latest known 0.19.1Raspberry Pi 5 Model B Rev 1.1 · Broadcom BCM2712 · litert-lm 0.16.1 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-09-01 | CARD.md · card.json | — |
nemotron-3-diarization__nemotron3_diar_encoder_low_latency_fp16 | voice-activity-detection | gpu_mldrift @ litert 2.2.0: 93.5% · 191 partition(s) | pass · 5593.795 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-09-25 | pass · 53.978 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-09-25 | galaxy-s26 gpu_mldrift: output_mismatch · 106.1672395 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (SDK 36) · 2026-09-24galaxy-s26 gpu_mldrift: pass · 136.183932 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (SDK 36) · 2026-09-24galaxy-s26 gpu_mldrift: output_mismatch · 80.3983855 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (SDK 36) · 2026-09-24 | CARD.md · card.json | run it |
nemotron-3-diarization__nemotron3_diar_encoder_offline_fp16 | voice-activity-detection | gpu_mldrift @ litert 2.2.0: 93.5% · 191 partition(s) | pass · 7200.502 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-09-25 | pass · 58.1 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-09-25 | galaxy-s26 gpu_mldrift: pass · 172.6847135 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (SDK 36) · 2026-09-24galaxy-s26 gpu_mldrift: output_mismatch · 87.179922 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (SDK 36) · 2026-09-24 | CARD.md · card.json | run it |
nemotron-3-diarization__nemotron3_diar_frontend | voice-activity-detection | gpu_mldrift @ litert 2.2.0: 100% · 1 partition(s) | pass · 0.425 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-09-25 | pass · 0.532 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-09-25 | galaxy-s26 gpu_mldrift: pass · 0.2066925 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (SDK 36) · 2026-09-24galaxy-s26 gpu_mldrift: output_mismatch · 0.2169005 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (SDK 36) · 2026-09-24 | CARD.md · card.json | run it |
nemotron-3-nano-30b-a3b-int4-ekv2048 | text-generation | — | not measured | not measured | mac-studio-m4-max cpu: passMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.1 · macOS 27.0 (26A428) · 2026-11-12mac-studio-m4-max cpu: passMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.1 · macOS 27.0 (26A428) · 2026-11-12mac-studio-m4-max cpu: pass · 12.76 tok/s decodeMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.1 · macOS 27.0 (26A428) · 2026-11-12mac-studio-m4-max gpu: passMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.1 · macOS 27.0 (26A428) · 2026-11-12mac-studio-m4-max gpu: passMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.1 · macOS 27.0 (26A428) · 2026-11-12mac-studio-m4-max gpu: pass · 75.22 tok/s decodeMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.1 · macOS 27.0 (26A428) · 2026-11-12 | CARD.md · card.json | — |
nemotron-3-nano-30b-a3b-int4-ekv4096 | text-generation | — | not measured | not measured | mac-studio-m4-max cpu: passMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.1 · macOS 27.0 (26A428) · 2026-11-12mac-studio-m4-max cpu: passMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.1 · macOS 27.0 (26A428) · 2026-11-12mac-studio-m4-max cpu: pass · 12.16 tok/s decodeMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.1 · macOS 27.0 (26A428) · 2026-11-12mac-studio-m4-max gpu: passMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.1 · macOS 27.0 (26A428) · 2026-11-12mac-studio-m4-max gpu: passMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.1 · macOS 27.0 (26A428) · 2026-11-12mac-studio-m4-max gpu: pass · 70.657 tok/s decodeMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.1 · macOS 27.0 (26A428) · 2026-11-12 | CARD.md · card.json | — |
nemotron-3-nano-30b-a3b-mixed-int4-ekv2048 | text-generation | — | not measured | not measured | mac-studio-m4-max cpu: passMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.1 · macOS 27.0 (26A428) · 2026-11-10mac-studio-m4-max cpu: passMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.1 · macOS 27.0 (26A428) · 2026-11-10mac-studio-m4-max cpu: pass · 8 tok/s decodeMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.1 · macOS 27.0 (26A428) · 2026-11-12mac-studio-m4-max gpu: passMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.1 · macOS 27.0 (26A428) · 2026-11-10mac-studio-m4-max gpu: passMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.1 · macOS 27.0 (26A428) · 2026-11-10mac-studio-m4-max gpu: pass · 71.387 tok/s decodeMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.1 · macOS 27.0 (26A428) · 2026-11-12 | CARD.md · card.json | — |
nemotron-3-nano-30b-a3b-mixed-int4-ekv4096 | text-generation | — | not measured | not measured | mac-studio-m4-max cpu: passMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.1 · macOS 27.0 (26A428) · 2026-11-10mac-studio-m4-max cpu: passMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.1 · macOS 27.0 (26A428) · 2026-11-10mac-studio-m4-max cpu: pass · 7.98 tok/s decodeMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.1 · macOS 27.0 (26A428) · 2026-11-12mac-studio-m4-max gpu: passMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.1 · macOS 27.0 (26A428) · 2026-11-10mac-studio-m4-max gpu: passMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.1 · macOS 27.0 (26A428) · 2026-11-10mac-studio-m4-max gpu: pass · 67.63 tok/s decodeMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.1 · macOS 27.0 (26A428) · 2026-11-12 | CARD.md · card.json | — |
nemotron-3-nano-4b-int8 | text-generation | — | not measured | not measured | galaxy-s26 cpu: fallbackGalaxy S26 (SM-S942Q) · SM8850 · litert-lm selfbuilt-b2f686e2 · Android 16 (BP4A.251205.006) · 2026-10-13galaxy-s26 cpu: fallbackGalaxy S26 (SM-S942Q) · SM8850 · litert-lm selfbuilt-b2f686e2 · Android 16 (BP4A.251205.006) · 2026-10-13galaxy-s26 cpu: fallbackGalaxy S26 (SM-S942Q) · SM8850 · litert-lm selfbuilt-b2f686e2 · Android 16 (BP4A.251205.006) · 2026-10-13galaxy-s26 cpu: fallback · 12.87 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-05galaxy-s26 gpu: run_failed stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-05mac-studio-m4-max cpu: pass stale — latest known 0.19.1Mac Studio (M4 Max) · litert-lm 0.18.0 · 2026-10-13mac-studio-m4-max cpu: pass stale — latest known 0.19.1Mac Studio (M4 Max) · litert-lm 0.16.0 · 2026-10-13mac-studio-m4-max cpu: pass stale — latest known 0.19.1Mac Studio (M4 Max) · litert-lm 0.18.0 · 2026-10-13mac-studio-m4-max cpu: pass stale — latest known 0.19.1Mac Studio (M4 Max) · litert-lm 0.18.0 · 2026-10-13mac-studio-m4-max cpu: pass · 22.73 tok/s decode stale — latest known 0.19.1mac-studio-m4-max · Apple M4 Max · litert-lm 0.16.0 · 2026-08-27mac-studio-m4-max gpu: pass stale — latest known 0.19.1Mac Studio (M4 Max) · litert-lm 0.18.0 · 2026-10-13mac-studio-m4-max gpu: pass · 83.3 tok/s decode stale — latest known 0.19.1mac-studio-m4-max · Apple M4 Max · litert-lm 0.16.0 · 2026-08-27raspberry-pi-5 cpu: pass · 2.2 tok/s decode stale — latest known 0.19.1Raspberry Pi 5 Model B Rev 1.1 · Broadcom BCM2712 · litert-lm 0.16.1 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-09-02 | CARD.md · card.json | — |
nemotron-h-4b | text-generation | — | not measured | not measured | raspberry-pi-5 cpu: pass · 1.97 tok/s decode stale — latest known 0.19.1Raspberry Pi 5 Model B Rev 1.1 · Broadcom BCM2712 · litert-lm 0.16.1 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-09-01 | CARD.md · card.json | — |
nemotron-labs-dory1.0-4b-int4 | text-generation | — | not measured | not measured | galaxy-s26 gpu: pass stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.1 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-10-24galaxy-s26 gpu: pass stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.1 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-10-24galaxy-s26 gpu: pass stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.1 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-10-26galaxy-s26 gpu: pass stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.1 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-10-26galaxy-s26 gpu: pass stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.1 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-10-26galaxy-s26 gpu: pass stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.1 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-10-26galaxy-s26 gpu: pass stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.1 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-10-22galaxy-s26 gpu: pass stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.1 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-10-22galaxy-s26 gpu: pass · 12.09 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.18.0 · Android 16 (BP4A.251205.006) · 2026-10-22mac-studio-m4-max gpu: pass · 57.705 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.18.0 · macOS 27.0 · 2026-10-22 | CARD.md · card.json | — |
nemotron-labs-dory1.0-4b-int4-1loop | text-generation | — | not measured | not measured | galaxy-s26 gpu: pass stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.1 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-10-24galaxy-s26 gpu: pass stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.1 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-10-24galaxy-s26 gpu: pass stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.1 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-10-26galaxy-s26 gpu: pass stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.1 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-10-26galaxy-s26 gpu: pass stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.1 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-10-26galaxy-s26 gpu: pass stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.1 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-10-26galaxy-s26 gpu: pass stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.1 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-10-23galaxy-s26 gpu: pass · 17.3225 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.18.0 · Android 16 (BP4A.251205.006) · 2026-10-23mac-studio-m4-max gpu: pass · 88.175 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.18.0 · macOS 27.0 · 2026-10-23 | CARD.md · card.json | — |
nemotron-labs-dory1.0-4b-int8 | text-generation | — | not measured | not measured | galaxy-s26 cpu: fallback stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.1 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-10-22galaxy-s26 cpu: fallback · 5.0625 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.18.0 · Android 16 (BP4A.251205.006) · 2026-10-22mac-studio-m4-max cpu: pass · 15.7 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.18.0 · macOS 27.0 · 2026-10-22pixel-8a cpu: fallback · 1.18 tok/s decode stale — latest known 0.19.1Pixel 8a · Google Tensor G3 · litert-lm 0.18.0 · Android 16 (CP1A.260505.005) · 2026-10-22 | CARD.md · card.json | — |
nemotron-nano-9b-v2-int4 | text-generation | — | not measured | not measured | mac-studio-m4-max cpu: pass · 13.985 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.18.0 · macOS 27.0 (26A428) · 2026-10-15mac-studio-m4-max gpu: pass · 45.52 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.18.0 · macOS 27.0 (26A428) · 2026-10-15pixel-8a gpu: load_failedPixel 8a · Tensor G3 · litert-lm prebuilt-adac974b · Android 16 (CP1A.260505.005) · 2026-10-15 | CARD.md · card.json | — |
nemotron-nano-9b-v2-int8 | text-generation | — | not measured | not measured | mac-studio-m4-max cpu: pass · 12.39 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.18.0 · macOS 27.0 (26A428) · 2026-10-15mac-studio-m4-max gpu: pass stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.17.1 · macOS 27.0 (26A428) · 2026-10-14mac-studio-m4-max gpu: pass · 41.02 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.18.0 · macOS 27.0 (26A428) · 2026-10-15 | CARD.md · card.json | — |
north-micro-vision-instruct-wi8 | image-text-to-text | — | not measured | not measured | galaxy-s26 gpu: pass · 10.4 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-08-25iphone-17-pro gpu: pass · 13.64 tok/s decode stale — latest known 0.19.1iPhone 17 Pro · litert-lm 0.16.0 · iOS 27.0 · 2026-08-19pixel-8a gpu: pass · 4.3 tok/s decode stale — latest known 0.19.1Pixel 8a · Tensor G3 · litert-lm 0.16.1 · 2026-08-19 | CARD.md · card.json | — |
oc-e-non-v1-int8-ekv4096 | text-generation | — | not measured | not measured | galaxy-s26 cpu: fallback · 7.4475 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.19.0 · Android 16 (BP4A.251205.006) · 2026-11-15galaxy-s26 gpu: pass · 12.77 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.19.0 · Android 16 (BP4A.251205.006) · 2026-11-15mac-studio-m4-max cpu: passMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.2 · macOS 27.0 (26A428) · 2026-11-14mac-studio-m4-max gpu: passMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.2 · macOS 27.0 (26A428) · 2026-11-14mac-studio-m4-max gpu: passMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.2 · macOS 27.0 (26A428) · 2026-11-14pixel-11-pro cpu: fallback stale — latest known 0.19.1Pixel 11 Pro · Google Tensor G6 · litert-lm 0.19.0 · Android 17 (CD1A.260905.001.B1) · 2026-11-14pixel-11-pro gpu: pass stale — latest known 0.19.1Pixel 11 Pro · Google Tensor G6 · litert-lm 0.19.0 · Android 17 (CD1A.260905.001.B1) · 2026-11-14 | CARD.md · card.json | — |
oc-e-non-v1-mixed-int4b32-fp16act-ekv4096 | text-generation | — | not measured | not measured | galaxy-s26 cpu: fallback · 9.125 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.19.0 · Android 16 (BP4A.251205.006) · 2026-11-15galaxy-s26 gpu: pass stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.19.0 · Android 16 (BP4A.251205.006) · 2026-11-15galaxy-s26 gpu: pass · 18.105 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.19.0 · Android 16 (BP4A.251205.006) · 2026-11-15mac-studio-m4-max cpu: passMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.2 · macOS 27.0 (26A428) · 2026-11-15mac-studio-m4-max gpu: passMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.2 · macOS 27.0 (26A428) · 2026-11-15mac-studio-m4-max gpu: passMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.2 · macOS 27.0 (26A428) · 2026-11-15pixel-11-pro cpu: fallback stale — latest known 0.19.1Pixel 11 Pro · Google Tensor G6 · litert-lm 0.19.0 · Android 17 (CD1A.260905.001.B1) · 2026-11-15pixel-11-pro gpu: pass stale — latest known 0.19.1Pixel 11 Pro · Google Tensor G6 · litert-lm 0.19.0 · Android 17 (CD1A.260905.001.B1) · 2026-11-15 | CARD.md · card.json | — |
olmo-2-1b-instruct-q4-block32-ekv4096 | text-generation | — | not measured | not measured | galaxy-s26 cpu: fallback · 12.1 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-05galaxy-s26 gpu: pass · 22.97 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-08-24raspberry-pi-5 cpu: pass · 2.13 tok/s decode stale — latest known 0.19.1Raspberry Pi 5 Model B Rev 1.1 · Broadcom BCM2712 · litert-lm 0.16.1 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-09-01 | CARD.md · card.json | — |
olmo-3-7b-think-int4 | text-generation | — | not measured | not measured | mac-studio-m4-max cpu: pass stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.17.1 · macOS 27.0 (26A428) · 2026-10-16mac-studio-m4-max cpu: pass · 7.105 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.18.0 · macOS 27.0 (26A428) · 2026-10-16mac-studio-m4-max gpu: pass · 34.76 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.18.0 · macOS 27.0 (26A428) · 2026-10-16 | CARD.md · card.json | — |
olmo-3-7b-think-int4-c2048 | text-generation | — | not measured | not measured | mac-studio-m4-max gpu: pass · 46.575 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.18.0 · macOS 27.0 (26A428) · 2026-10-16 | CARD.md · card.json | — |
olmo-3-7b-think-int8 | text-generation | — | not measured | not measured | mac-studio-m4-max cpu: pass stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.17.1 · macOS 27.0 (26A428) · 2026-10-16mac-studio-m4-max cpu: pass · 6.695 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.18.0 · macOS 27.0 (26A428) · 2026-10-16mac-studio-m4-max gpu: pass · 30.395 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.18.0 · macOS 27.0 (26A428) · 2026-10-16 | CARD.md · card.json | — |
open-decision-deberta-v3-large__deberta_v3_large_decision_s256_wfp16 | typed decisions (state + choice / score / noul questions -> one calibrated distribution per question; text-classification) | gpu_mldrift @ litert 2.2.0: 89.1% · 196 partition(s) | run_failed stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-10-02 | load_failed stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-10-02 | galaxy-s26 cpu_xnnpack: pass · 631.75276 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (build BP4A.251205.006.S942QOPS1AZH9) · 2026-10-03galaxy-s26 gpu_mldrift: pass · 298.967292 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (SDK 36, build BP4A.251205.006) · 2026-10-02galaxy-s26 gpu_mldrift: pass · 407.64333350000004 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (SDK 36, build BP4A.251205.006) · 2026-10-02galaxy-s26 gpu_mldrift: pass · 185.282708 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (build BP4A.251205.006.S942QOPS1AZH9) · 2026-10-04galaxy-s26 npu_qnn: output_mismatch · 129.858073 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (SDK 36, build BP4A.251205.006) · 2026-10-02 | CARD.md · card.json | — |
open-decision-deberta-v3-large__deberta_v3_large_decision_s512_wfp16 | typed decisions (state + choice / score / noul questions -> one calibrated distribution per question; text-classification) | gpu_mldrift @ litert 2.2.0: 89.1% · 196 partition(s) | run_failed stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-10-02 | load_failed stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-10-02 | galaxy-s26 gpu_mldrift: pass · 619.77177 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (SDK 36, build BP4A.251205.006) · 2026-10-02galaxy-s26 gpu_mldrift: pass · 697.48927 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (SDK 36, build BP4A.251205.006) · 2026-10-02galaxy-s26 gpu_mldrift: pass · 1027.440625 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (build BP4A.251205.006.S942QOPS1AZH9) · 2026-10-04galaxy-s26 gpu_mldrift: pass stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (build BP4A.251205.006.S942QOPS1AZH9) · 2026-10-05galaxy-s26 gpu_mldrift: pass · 1031.236406 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (build BP4A.251205.006.S942QOPS1AZH9) · 2026-10-04galaxy-s26 npu_qnn: load_failed stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (SDK 36, build BP4A.251205.006) · 2026-10-02 | CARD.md · card.json | — |
ormbg | image-segmentation | — | pass · 4525.905 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | pass · 65.155 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | raspberry-pi-5 cpu_xnnpack: pass · 2989.129 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
ornith-1.0-9b-int8 | text-generation | — | not measured | not measured | mac-studio-m4-max cpu: pass stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.17.1 · macOS 27.0 (26A428) · 2026-10-15mac-studio-m4-max cpu: pass · 15.17 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.18.0 · macOS 27.0 (26A428) · 2026-10-15mac-studio-m4-max gpu: pass · 39.99 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.18.0 · macOS 27.0 (26A428) · 2026-10-15 | CARD.md · card.json | — |
ornith-1.0-9b-mixed-int4 | text-generation | — | not measured | not measured | mac-studio-m4-max cpu: pass stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.17.1 · macOS 27.0 (26A428) · 2026-10-15mac-studio-m4-max cpu: pass · 15.765 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.18.0 · macOS 27.0 (26A428) · 2026-10-15mac-studio-m4-max gpu: pass · 44.925 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.18.0 · macOS 27.0 (26A428) · 2026-10-15 | CARD.md · card.json | — |
ornith-1.5-9b-int8 | text-generation | — | not measured | not measured | mac-studio-m4-max cpu: pass stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.17.1 · macOS 27.0 (26A428) · 2026-10-15mac-studio-m4-max cpu: pass · 14.975 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.18.0 · macOS 27.0 (26A428) · 2026-10-15mac-studio-m4-max gpu: pass · 39.565 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.18.0 · macOS 27.0 (26A428) · 2026-10-15 | CARD.md · card.json | — |
ornith-1.5-9b-mixed-int4 | text-generation | — | not measured | not measured | mac-studio-m4-max cpu: pass stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.17.1 · macOS 27.0 (26A428) · 2026-10-15mac-studio-m4-max cpu: pass · 14.78 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.18.0 · macOS 27.0 (26A428) · 2026-10-15mac-studio-m4-max gpu: pass · 45.58 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.18.0 · macOS 27.0 (26A428) · 2026-10-15 | CARD.md · card.json | — |
ovis2_5-2b | image-text-to-text | — | not measured | not measured | galaxy-s26 gpu: pass · 26.83 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-08-25 | CARD.md · card.json | — |
ovisocr2-int8 | image-text-to-text | — | not measured | not measured | galaxy-s26 cpu: run_failed stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-05galaxy-s26 gpu: pass · 35.18 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-05iphone-17-pro cpu: pass · 18.05 tok/s decode stale — latest known 0.19.1iPhone 17 Pro · A19-Pro · litert-lm 0.16.0 · iOS 27.0 · 2026-09-01iphone-17-pro gpu: pass · 48.5 tok/s decode stale — latest known 0.19.1iPhone 17 Pro · A19-Pro · litert-lm 0.16.0 · iOS 27.0 · 2026-09-01mac-studio-m4-max cpu: pass · 48.93 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · litert-lm 0.16.0 · 2026-09-01mac-studio-m4-max gpu: pass · 140.9 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · litert-lm 0.16.0 · 2026-09-01pixel-8a cpu: run_failed stale — latest known 0.19.1Pixel 8a · Tensor G3 · litert-lm 0.16.0 · Android 16 · 2026-09-05 | CARD.md · card.json | — |
paddleocr-vl-1.6 | image-text-to-text | — | not measured | not measured | galaxy-s26 cpu: run_failed stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-05galaxy-s26 gpu: pass · 35.59 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-08-25pixel-8a cpu: run_failed stale — latest known 0.19.1Pixel 8a · Tensor G3 · litert-lm 0.16.0 · Android 16 · 2026-09-05 | CARD.md · card.json | — |
panns-cnn14-audioset | audio-classification | — | pass · 2473.497 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | pass · 13.235 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | galaxy-s26 gpu_mldrift: pass · 12.154 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 · 2026-08-26galaxy-s26 npu_qnn: pass · 3.888 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · QAIRT (Hexagon, JIT on-device) · Android 16 · 2026-08-26raspberry-pi-5 cpu_xnnpack: pass · 390.821 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
parakeet-tdt-0.6b-v3__parakeet_tdt_0.6b_v3_30s_i8_stateful | automatic-speech-recognition | — | pass · 7451.368 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-09-25 | load_failed stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-09-25 | galaxy-s26 cpu_xnnpack: fallback stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU, Hexagon HTP) · litert 2.2.0 · Android 16 (BP4A.251205.006.S942QOPS1AZH9, security patch 2026-08-05; the run header's device block) · 2026-10-19galaxy-s26 gpu_mldrift: pass stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU, Hexagon HTP) · litert 2.2.0 · Android 16 (BP4A.251205.006.S942QOPS1AZH9, security patch 2026-08-05; the run header's device block) · 2026-10-19mac-studio-m4-max cpu: passMac Studio (Apple M4 Max, 128 GB) · Apple M4 Max · litert-lm selfbuilt-66058c82+omni-eval-patches · macOS 27.0 (26A428) · 2026-09-25 | CARD.md · card.json | run it |
parakeet-tdt_ctc-0.6b-ja__parakeet_tdt_ctc_0.6b_ja_5s_f32 | automatic-speech-recognition | — | load_failed stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-13 | load_failed stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-13 | galaxy-s26 gpu_mldrift: pass · 63.606 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 · 2026-08-27galaxy-s26 npu_qnn: pass · 30.326 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · QAIRT (Hexagon, AOT host-compiled, SM8850 target) · Android 16 · 2026-08-27 | CARD.md · card.json | — |
parakeet-tdt_ctc-0.6b-ja__parakeet_tdt_ctc_0.6b_ja_5s_f32_stateful | automatic-speech-recognition | — | not measured | not measured | galaxy-s26 gpu_mldrift: pass · 63.033 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 · 2026-08-27galaxy-s26 npu_qnn: pass · 30.315 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · QAIRT (Hexagon, AOT host-compiled, SM8850 target) · Android 16 · 2026-08-27 | CARD.md · card.json | — |
parakeet-tdt_ctc-0.6b-ja__parakeet_tdt_ctc_0.6b_ja_5s_i8 | automatic-speech-recognition | — | pass · 1184.507 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-13 | load_failed stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-13 | galaxy-s26 gpu_mldrift: load_failed stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 · 2026-08-27galaxy-s26 npu_qnn: load_failed stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · QAIRT (Hexagon, AOT host-compiled, SM8850 target) · Android 16 · 2026-08-27raspberry-pi-5 cpu_xnnpack: pass · 110.356 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31raspberry-pi-5 cpu_xnnpack: pass · 275.894 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
parakeet-tdt_ctc-0.6b-ja__parakeet_tdt_ctc_0.6b_ja_5s_i8_stateful | automatic-speech-recognition | — | not measured | not measured | galaxy-s26 gpu_mldrift: load_failed stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 · 2026-08-27galaxy-s26 npu_qnn: load_failed stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · QAIRT (Hexagon, AOT host-compiled, SM8850 target) · Android 16 · 2026-08-27raspberry-pi-5 cpu_xnnpack: pass · 6.894 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31raspberry-pi-5 cpu_xnnpack: pass · 2.033 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31raspberry-pi-5 cpu_xnnpack: pass · 273.607 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | — |
phi4-mini-reasoning | text-generation | — | not measured | not measured | galaxy-s26 cpu: fallback · 7.04 tok/s decodeGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm prebuilt-adac974b · Android 16 · 2026-10-05galaxy-s26 gpu: pass · 13.11 tok/s decodeGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm prebuilt-adac974b · Android 16 · 2026-10-05galaxy-s26 npu: load_failedGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm selfbuilt-3dbb23e1e · QAIRT 2.47.0.260601114230 (QNN JIT libraries) · Android 16 · 2026-10-05mac-studio-m4-max cpu: fallback · 15.34 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.17.1 · 2026-10-05mac-studio-m4-max cpu: fallback · 19.73 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.17.1 · 2026-10-05mac-studio-m4-max gpu: pass · 105.61 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.17.1 · 2026-10-05mac-studio-m4-max gpu: pass · 114.9 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.17.1 · 2026-10-05raspberry-pi-5 cpu: pass · 1.62 tok/s decode stale — latest known 0.19.1Raspberry Pi 5 Model B Rev 1.1 · Broadcom BCM2712 · litert-lm 0.16.1 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-09-01 | CARD.md · card.json | — |
phocinae-largha-150m__phocinae_largha_150m_l256_fp16fc_embeds | text-classification | gpu_mldrift @ litert 2.2.0: 91.7% · 150 partition(s) | pass · 454.525 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-10-11 | pass · 17.063 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-10-11 | galaxy-s26 cpu_xnnpack: output_mismatch · 38.767083 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 36 build BP4A.251205.006.S942QOPS1AZH9 · 2026-10-10galaxy-s26 gpu_mldrift: output_mismatch · 15.331458 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 36 build BP4A.251205.006.S942QOPS1AZH9 · 2026-10-10galaxy-s26 gpu_mldrift: output_mismatch · 20.357761 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 36 build BP4A.251205.006.S942QOPS1AZH9 · 2026-10-10galaxy-s26 gpu_mldrift: output_mismatch · 28.867031 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 36 build BP4A.251205.006.S942QOPS1AZH9 · 2026-10-10galaxy-s26 npu_qnn: output_mismatch · 16.793229 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 36 build BP4A.251205.006.S942QOPS1AZH9 · 2026-10-10 | CARD.md · card.json | run it |
phocinae-largha-150m__phocinae_largha_150m_l256_fp16fc_int8emb | text-classification | gpu_mldrift @ litert 2.2.0: 91.6% · 150 partition(s) | pass · 450.27 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-10-11 | pass · 16.978 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-10-11 | galaxy-s26 cpu_xnnpack: output_mismatch · 38.96224 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 36 build BP4A.251205.006.S942QOPS1AZH9 · 2026-10-10galaxy-s26 gpu_mldrift: output_mismatch · 15.82375 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 36 build BP4A.251205.006.S942QOPS1AZH9 · 2026-10-10galaxy-s26 gpu_mldrift: output_mismatch · 20.815052 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 36 build BP4A.251205.006.S942QOPS1AZH9 · 2026-10-10galaxy-s26 gpu_mldrift: output_mismatch · 26.851146 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 36 build BP4A.251205.006.S942QOPS1AZH9 · 2026-10-10galaxy-s26 npu_qnn: output_mismatch · 17.514896 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 36 build BP4A.251205.006.S942QOPS1AZH9 · 2026-10-10 | CARD.md · card.json | run it |
phocinae-largha-150m__phocinae_largha_150m_l384_fp16fc_embeds | text-classification | gpu_mldrift @ litert 2.2.0: 91.7% · 150 partition(s) | pass · 701.555 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-10-11 | pass · 20.103 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-10-11 | galaxy-s26 gpu_mldrift: output_mismatch · 22.439323 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 36 build BP4A.251205.006.S942QOPS1AZH9 · 2026-10-10galaxy-s26 gpu_mldrift: output_mismatch · 22.93 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (SDK 36) · 2026-10-11galaxy-s26 gpu_mldrift: output_mismatch · 29.996458 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 36 build BP4A.251205.006.S942QOPS1AZH9 · 2026-10-10galaxy-s26 gpu_mldrift: output_mismatch · 68.76 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (SDK 36) · 2026-10-11galaxy-s26 npu_qnn: output_mismatch · 19.69052 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 36 build BP4A.251205.006.S942QOPS1AZH9 · 2026-10-10galaxy-s26 npu_qnn: output_mismatch · 19.77 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 (SDK 36) · 2026-10-11 | CARD.md · card.json | run it |
phocinae-largha-150m__phocinae_largha_150m_l384_fp16fc_int8emb | text-classification | gpu_mldrift @ litert 2.2.0: 91.6% · 150 partition(s) | pass · 726.63 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-10-11 | pass · 20.145 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-10-11 | galaxy-s26 cpu_xnnpack: output_mismatch · 63.004531 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 36 build BP4A.251205.006.S942QOPS1AZH9 · 2026-10-10galaxy-s26 gpu_mldrift: output_mismatch · 50.082344 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 36 build BP4A.251205.006.S942QOPS1AZH9 · 2026-10-10galaxy-s26 gpu_mldrift: output_mismatch · 55.495885 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 36 build BP4A.251205.006.S942QOPS1AZH9 · 2026-10-10galaxy-s26 gpu_mldrift: output_mismatch · 42.128542 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 36 build BP4A.251205.006.S942QOPS1AZH9 · 2026-10-10 | CARD.md · card.json | run it |
phocinae-largha-150m__phocinae_largha_150m_l512_fp16fc_embeds | text-classification | gpu_mldrift @ litert 2.2.0: 91.7% · 150 partition(s) | pass · 983.827 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-10-11 | pass · 24.285 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-10-11 | galaxy-s26 gpu_mldrift: output_mismatch · 30.739948 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 36 build BP4A.251205.006.S942QOPS1AZH9 · 2026-10-10galaxy-s26 gpu_mldrift: output_mismatch · 40.445521 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 36 build BP4A.251205.006.S942QOPS1AZH9 · 2026-10-10galaxy-s26 npu_qnn: output_mismatch · 26.921459 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 36 build BP4A.251205.006.S942QOPS1AZH9 · 2026-10-10 | CARD.md · card.json | run it |
phocinae-largha-150m__phocinae_largha_150m_l512_fp16fc_int8emb | text-classification | gpu_mldrift @ litert 2.2.0: 91.6% · 150 partition(s) | pass · 976.315 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-10-11 | pass · 24.563 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-10-11 | galaxy-s26 cpu_xnnpack: output_mismatch · 96.553698 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 36 build BP4A.251205.006.S942QOPS1AZH9 · 2026-10-10galaxy-s26 gpu_mldrift: output_mismatch · 62.727917 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 36 build BP4A.251205.006.S942QOPS1AZH9 · 2026-10-10 | CARD.md · card.json | run it |
pidnet-s-cityscapes | image-segmentation | — | pass · 670.503 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | output_mismatch · 14.763 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | raspberry-pi-5 cpu_xnnpack: pass · 482.334 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
places365-resnet18 | image-classification | — | pass · 226.743 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | pass · 2.723 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | galaxy-s26 gpu_mldrift: pass · 3.489 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 · 2026-08-26galaxy-s26 npu_qnn: pass · 0.918 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · QAIRT (Hexagon, JIT on-device) · Android 16 · 2026-08-26raspberry-pi-5 cpu_xnnpack: pass · 33.995 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
plantnet-300k-resnet18 | image-classification | — | pass · 52.538 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | pass · 2.665 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | raspberry-pi-5 cpu_xnnpack: pass · 34.147 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
polaris-4b | text-generation | — | not measured | not measured | mac-studio-m4-max gpu: pass · 68.87 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.14.0 · 2026-07-23raspberry-pi-5 cpu: pass · 1.49 tok/s decode stale — latest known 0.19.1Raspberry Pi 5 Model B Rev 1.1 · Broadcom BCM2712 · litert-lm 0.16.1 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-09-02 | CARD.md · card.json | — |
pp-ocrv5__ppocr_det_fp16 | text-detection | — | pass · 916.36 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | pass · 12.325 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | galaxy-s26 gpu_mldrift: pass · 17.59 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 · 2026-08-25galaxy-s26 npu_qnn: pass · 8.634 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · QAIRT (Hexagon, JIT on-device) · Android 16 · 2026-08-25raspberry-pi-5 cpu_xnnpack: pass · 460.168 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
pp-ocrv5__ppocr_rec_fp16 | text-recognition | — | pass · 449.3 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | output_mismatch · 15.125 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | galaxy-s26 gpu_mldrift: pass · 7.998 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 · 2026-08-26galaxy-s26 npu_qnn: pass · 5.911 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · QAIRT (Hexagon, JIT on-device) · Android 16 · 2026-08-26raspberry-pi-5 cpu_xnnpack: pass · 86.447 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
pp-ocrv6-small__ppocrv6_small_det_640_fp32 | image-to-text | gpu_mldrift @ litert 2.2.0: 4.5% · 13 partition(s) | pass · 214.418 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-09-25 | pass · 10.045 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-09-25 | not measured | CARD.md · card.json | run it |
pp-ocrv6-small__ppocrv6_small_rec_320_fp32 | image-to-text | gpu_mldrift @ litert 2.2.0: 4.6% · 13 partition(s) | pass · 43.815 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-09-25 | output_mismatch · 13.438 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-09-25 | not measured | CARD.md · card.json | run it |
pp-ocrv6-small__ppocrv6_small_rec_640_fp32 | image-to-text | gpu_mldrift @ litert 2.2.0: 4.6% · 13 partition(s) | pass · 83.455 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-09-25 | pass · 23.417 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-09-25 | not measured | CARD.md · card.json | run it |
pp-ocrv6-small__ppocrv6_small_rec_960_fp32 | image-to-text | gpu_mldrift @ litert 2.2.0: 4.6% · 13 partition(s) | pass · 125.062 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-09-25 | pass · 34.363 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-09-25 | not measured | CARD.md · card.json | run it |
qwen25-coder-1.5b-int4 | text-generation | — | not measured | not measured | galaxy-s26 cpu: fallback · 29.73 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-05galaxy-s26 gpu: pass · 10.07 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-08-24pixel-8a cpu: pass · 12.57 tok/s decode stale — latest known 0.19.1Pixel 8a · Tensor G3 · litert-lm 0.16.0 · 2026-08-19pixel-8a gpu: pass · 12 tok/s decode stale — latest known 0.19.1Pixel 8a · Tensor G3 · litert-lm 0.16.0 · 2026-08-19raspberry-pi-5 cpu: pass · 5.61 tok/s decode stale — latest known 0.19.1Raspberry Pi 5 Model B Rev 1.1 · Broadcom BCM2712 · litert-lm 0.16.1 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-09-01 | CARD.md · card.json | — |
qwen3-30b-a3b-mixed-int4-ekv2048 | text-generation | — | not measured | not measured | mac-studio-m4-max cpu: run_failed stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.0 · macOS 27.0 (26A428) · 2026-11-03mac-studio-m4-max gpu: load_failed stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.0 · macOS 27.0 (26A428) · 2026-11-03 | CARD.md · card.json | — |
qwen3-30b-a3b-mixed-int4-ekv4096 | text-generation | — | not measured | not measured | mac-studio-m4-max cpu: run_failed stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.0 · macOS 27.0 (26A428) · 2026-11-03mac-studio-m4-max gpu: load_failed stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.0 · macOS 27.0 (26A428) · 2026-11-03 | CARD.md · card.json | — |
qwen3-32b-int8-ekv4096 | text-generation | — | not measured | not measured | mac-studio-m4-max cpu: passMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.1 · macOS 27.0 (26A428) · 2026-11-09mac-studio-m4-max cpu: pass · 5.03 tok/s decodeMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.1 · macOS 27.0 (26A428) · 2026-11-09mac-studio-m4-max gpu: passMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.1 · macOS 27.0 (26A428) · 2026-11-09mac-studio-m4-max gpu: passMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.1 · macOS 27.0 (26A428) · 2026-11-09mac-studio-m4-max gpu: pass · 11.86 tok/s decodeMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.1 · macOS 27.0 (26A428) · 2026-11-09 | CARD.md · card.json | — |
qwen3-4b-thinking | text-generation | — | not measured | not measured | galaxy-s26 cpu: fallback · 8.78 tok/s decodeGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm prebuilt-adac974b · Android 16 · 2026-10-05galaxy-s26 cpu: fallback · 5.99 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-05galaxy-s26 cpu: fallback · 7.29 tok/s decodeGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm prebuilt-adac974b · Android 16 · 2026-10-05galaxy-s26 gpu: pass · 20.72 tok/s decodeGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm prebuilt-adac974b · Android 16 · 2026-10-05galaxy-s26 gpu: pass · 18.11 tok/s decodeGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm prebuilt-adac974b · Android 16 · 2026-10-05mac-studio-m4-max cpu: fallback · 16 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.17.1 · 2026-10-05mac-studio-m4-max cpu: fallback · 18.41 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.17.1 · 2026-10-05mac-studio-m4-max gpu: pass · 89.65 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.17.1 · 2026-10-05mac-studio-m4-max gpu: pass · 92.09 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.17.1 · 2026-10-05raspberry-pi-5 cpu: pass · 1.49 tok/s decode stale — latest known 0.19.1Raspberry Pi 5 Model B Rev 1.1 · Broadcom BCM2712 · litert-lm 0.16.1 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-09-01 | CARD.md · card.json | — |
qwen3-4b-thinking-dynamic-wi4b32-afp32 | text-generation | — | not measured | not measured | galaxy-s26 cpu: passGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm selfbuilt-b2f686e2 · Android 16 · 2026-10-07galaxy-s26 gpu: passGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm selfbuilt-b2f686e2 · Android 16 · 2026-10-07galaxy-s26 gpu: pass · 11.2 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-08-24raspberry-pi-5 cpu: pass · 1.43 tok/s decode stale — latest known 0.19.1Raspberry Pi 5 Model B Rev 1.1 · Broadcom BCM2712 · litert-lm 0.16.1 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-09-01 | CARD.md · card.json | — |
qwen3-asr-1.7b | automatic-speech-recognition | — | not measured | not measured | galaxy-s26 cpu: pass · 2490 ms stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.17.1 · Android 16 (build BP4A.251205.006.S942QOPS1AZH9) · 2026-10-09galaxy-s26 cpu: fallback · 25.6 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · SM8850 · litert-lm 0.18.0 · Android 16 (BP4A.251205.006) · 2026-10-13galaxy-s26 gpu: pass · 24.79 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · SM8850 · litert-lm 0.18.0 · Android 16 (BP4A.251205.006) · 2026-10-13mac-studio-m4-max cpu: pass · 1700.9 ms stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.17.1 · macOS 27.0 · 2026-10-03mac-studio-m4-max cpu: fallback · 1243.06 ms stale — latest known 0.19.1Mac Studio (M4 Max) · litert-lm 0.18.0 · 2026-10-13mac-studio-m4-max gpu: pass · 663.5 ms stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.17.1 · macOS 27.0 · 2026-10-03mac-studio-m4-max gpu: pass · 555.04 ms stale — latest known 0.19.1Mac Studio (M4 Max) · litert-lm 0.18.0 · 2026-10-13pixel-8a cpu: fallback · 11.04 tok/s decode stale — latest known 0.19.1Pixel 8a · Tensor G3 · litert-lm 0.18.0 · Android 16 (CP1A.260505.005) · 2026-10-13pixel-8a gpu: pass · 10.4 tok/s decode stale — latest known 0.19.1Pixel 8a · Tensor G3 · litert-lm 0.18.0 · Android 16 (CP1A.260505.005) · 2026-10-13 | CARD.md · card.json | — |
qwen3-coder-30b-a3b-mixed-int4-ekv2048 | text-generation | — | not measured | not measured | mac-studio-m4-max cpu: run_failedMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.1 · macOS 27.0 (26A428) · 2026-11-09mac-studio-m4-max gpu: load_failedMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.1 · macOS 27.0 (26A428) · 2026-11-09 | CARD.md · card.json | — |
qwen3-coder-30b-a3b-mixed-int4-ekv4096 | text-generation | — | not measured | not measured | mac-studio-m4-max cpu: run_failedMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.1 · macOS 27.0 (26A428) · 2026-11-09mac-studio-m4-max gpu: load_failedMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.1 · macOS 27.0 (26A428) · 2026-11-09 | CARD.md · card.json | — |
qwen3-embedding-0.6b | sentence-similarity | — | load_failed stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | load_failed stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | galaxy-s26 gpu_mldrift: pass · 44.738 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 · 2026-08-27galaxy-s26 npu_qnn: pass · 25.44 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · QAIRT (Hexagon, AOT host-compiled, SM8850 target) · Android 16 · 2026-08-27 | CARD.md · card.json | — |
qwen3-reranker-0.6b | text-ranking | — | load_failed stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | load_failed stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | galaxy-s26 gpu_mldrift: pass · 78.494 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 · 2026-08-27galaxy-s26 npu_qnn: pass · 47.508 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · QAIRT (Hexagon, AOT host-compiled, SM8850 target) · Android 16 · 2026-08-27 | CARD.md · card.json | — |
qwen35-0.8b | text-generation | — | not measured | not measured | galaxy-s26 cpu: fallback · 45.78 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · SM8850 · litert-lm 0.18.0 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-10-24galaxy-s26 gpu: pass · 30.52 tok/s decodeGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm prebuilt-adac974b · Android 16 · 2026-10-12galaxy-s26 gpu: pass · 28.61 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · SM8850 · litert-lm 0.18.0 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-10-24pixel-8a cpu: fallback · 20 tok/s decodePixel 8a · Tensor G3 · litert-lm prebuilt-adac974b · Android 16 · 2026-10-12pixel-8a gpu: load_failedPixel 8a · Tensor G3 · litert-lm prebuilt-adac974b · Android 16 · 2026-10-12raspberry-pi-5 cpu: pass · 7.06 tok/s decode stale — latest known 0.19.1Raspberry Pi 5 Model B Rev 1.1 · Broadcom BCM2712 · litert-lm 0.16.1 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-09-01 | CARD.md · card.json | — |
qwen35-0.8b-vl-int8 | image-text-to-text | — | not measured | not measured | galaxy-s26 cpu: fallback · 32.41 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-05galaxy-s26 gpu: pass · 35.07 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-05galaxy-s26 gpu: pass · 34.36 tok/s decodeGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm prebuilt-adac974b · Android 16 · 2026-10-12iphone-17-pro cpu: pass · 26.01 tok/s decode stale — latest known 0.19.1iPhone 17 Pro · A19-Pro · litert-lm 0.16.0 · iOS 27.0 · 2026-08-31iphone-17-pro gpu: pass · 45.71 tok/s decode stale — latest known 0.19.1iPhone 17 Pro · A19-Pro · litert-lm 0.16.0 · iOS 27.0 · 2026-08-28mac-studio-m4-max cpu: pass · 46.63 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · litert-lm 0.16.0 · 2026-08-28mac-studio-m4-max gpu: pass · 126.17 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · litert-lm 0.16.0 · 2026-08-28pixel-8a cpu: fallback · 15.77 tok/s decode stale — latest known 0.19.1Pixel 8a · Tensor G3 · litert-lm 0.16.0 · Android 16 · 2026-09-05pixel-8a cpu: fallback · 19.86 tok/s decodePixel 8a · Tensor G3 · litert-lm prebuilt-adac974b · Android 16 · 2026-10-12pixel-8a gpu: load_failedPixel 8a · Tensor G3 · litert-lm prebuilt-adac974b · Android 16 · 2026-10-12raspberry-pi-5 cpu: pass · 6.91 tok/s decode stale — latest known 0.19.1Raspberry Pi 5 Model B Rev 1.1 · Broadcom BCM2712 · litert-lm 0.16.1 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-09-01 | CARD.md · card.json | — |
qwen35-08b-v4gpufix | text-generation | — | not measured | not measured | pixel-8a gpu: pass · 7.45 tok/s decode stale — latest known 0.19.1pixel-8a · Google Tensor G3 · litert-lm 0.16.0 · 2026-08-13 | CARD.md · card.json | — |
qwen35-2b | text-generation | — | not measured | not measured | galaxy-s26 gpu: pass · 18.48 tok/s decodeGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm prebuilt-adac974b · Android 16 · 2026-10-13mac-m4-max cpu: pass · 37.6 tok/s decode stale — latest known 0.19.1mac-m4-max · litert-lm 0.16.0 · 2026-08-14mac-m4-max gpu: pass · 114.34 tok/s decode stale — latest known 0.19.1mac-m4-max · litert-lm 0.16.0 · 2026-08-14pixel-8a cpu: fallback · 9.69 tok/s decodePixel 8a · Tensor G3 · litert-lm prebuilt-adac974b · Android 16 · 2026-10-13pixel-8a gpu: load_failedPixel 8a · Tensor G3 · litert-lm prebuilt-adac974b · Android 16 · 2026-10-12raspberry-pi-5 cpu: pass · 4.25 tok/s decode stale — latest known 0.19.1Raspberry Pi 5 Model B Rev 1.1 · Broadcom BCM2712 · litert-lm 0.16.1 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-09-01 | CARD.md · card.json | — |
qwen35-2b-vl-int8 | image-text-to-text | — | not measured | not measured | galaxy-s26 cpu: fallback · 18.98 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-05galaxy-s26 gpu: pass · 17.47 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-05pixel-8a cpu: fallback · 10.54 tok/s decodePixel 8a · Tensor G3 · litert-lm prebuilt-adac974b · Android 16 · 2026-10-12pixel-8a gpu: load_failedPixel 8a · Tensor G3 · litert-lm prebuilt-adac974b · Android 16 · 2026-10-12raspberry-pi-5 cpu: pass · 4.24 tok/s decode stale — latest known 0.19.1Raspberry Pi 5 Model B Rev 1.1 · Broadcom BCM2712 · litert-lm 0.16.1 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-09-01 | CARD.md · card.json | — |
qwen35-4b | text-generation | — | not measured | not measured | galaxy-s26 cpu: fallbackGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm prebuilt-adac974b · Android 16 · 2026-10-13galaxy-s26 cpu: fallback · 8.93 tok/s decodeGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm prebuilt-adac974b · Android 16 · 2026-10-13mac-m4-max cpu: pass · 19.9 tok/s decode stale — latest known 0.19.1mac-m4-max · litert-lm 0.16.0 · 2026-08-14mac-m4-max gpu: pass · 61.98 tok/s decode stale — latest known 0.19.1mac-m4-max · litert-lm 0.16.0 · 2026-08-14mac-studio-m4-max cpu: pass stale — latest known 0.19.1Mac Studio (M4 Max) · litert-lm 0.18.0 · 2026-10-13mac-studio-m4-max gpu: pass stale — latest known 0.19.1Mac Studio (M4 Max) · litert-lm 0.18.0 · 2026-10-13pixel-8a cpu: fallbackPixel 8a · Tensor G3 · litert-lm prebuilt-adac974b · Android 16 · 2026-10-13raspberry-pi-5 cpu: pass · 1.83 tok/s decode stale — latest known 0.19.1Raspberry Pi 5 Model B Rev 1.1 · Broadcom BCM2712 · litert-lm 0.16.1 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-09-01 | CARD.md · card.json | — |
qwen35-4b-mixed-int4 | text-generation | — | not measured | not measured | galaxy-s26 cpu: passGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm selfbuilt-b2f686e2 · 2026-10-09galaxy-s26 gpu: pass · 5.39 tok/s decodeGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm prebuilt-adac974b · Android 16 · 2026-10-12iphone-17-pro cpu: pass · 8.87 tok/s decode stale — latest known 0.19.1iphone-17-pro · A19-Pro · litert-lm 0.16.0 · iOS 27.0 · 2026-08-27iphone-17-pro gpu: pass · 11.38 tok/s decode stale — latest known 0.19.1iphone-17-pro · A19-Pro · litert-lm 0.16.0 · iOS 27.0 · 2026-08-27mac-studio-m4-max cpu: pass · 20.07 tok/s decode stale — latest known 0.19.1mac-studio-m4-max · Apple M4 Max · litert-lm 0.16.0 · 2026-08-27mac-studio-m4-max gpu: pass · 68.45 tok/s decode stale — latest known 0.19.1mac-studio-m4-max · Apple M4 Max · litert-lm 0.16.0 · 2026-08-27pixel-8a cpu: fallback · 5.59 tok/s decodePixel 8a · Tensor G3 · litert-lm prebuilt-adac974b · Android 16 · 2026-10-12pixel-8a cpu: fallback · 6.3 tok/s decode stale — latest known 0.19.1pixel-8a · Tensor G3 · litert-lm 0.16.0 · Android 16 · 2026-08-27pixel-8a gpu: load_failedPixel 8a · Tensor G3 · litert-lm prebuilt-adac974b · Android 16 · 2026-10-12raspberry-pi-5 cpu: pass · 2.37 tok/s decode stale — latest known 0.19.1Raspberry Pi 5 Model B Rev 1.1 · Broadcom BCM2712 · litert-lm 0.16.1 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-09-01 | CARD.md · card.json | — |
qwen35-9b | text-generation | — | not measured | not measured | mac-studio-m4-max cpu: pass stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.18.0 · macOS 27.0 (26A428) · 2026-10-16mac-studio-m4-max cpu: pass stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.17.1 · macOS 27.0 (26A428) · 2026-10-16mac-studio-m4-max cpu: pass stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.18.0 · macOS 27.0 (26A428) · 2026-10-16mac-studio-m4-max cpu: pass stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.18.0 · macOS 27.0 (26A428) · 2026-10-16mac-studio-m4-max gpu: pass stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.18.0 · macOS 27.0 (26A428) · 2026-10-16mac-studio-m4-max gpu: pass stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.18.0 · macOS 27.0 (26A428) · 2026-10-16mac-studio-m4-max gpu: pass stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.18.0 · macOS 27.0 (26A428) · 2026-10-16 | CARD.md · card.json | — |
qwen35-9b-mixed-int4 | text-generation | — | not measured | not measured | mac-studio-m4-max cpu: pass stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.18.0 · macOS 27.0 (26A428) · 2026-10-16mac-studio-m4-max cpu: pass stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.17.1 · macOS 27.0 (26A428) · 2026-10-16mac-studio-m4-max cpu: pass stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.18.0 · macOS 27.0 (26A428) · 2026-10-16mac-studio-m4-max cpu: pass stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.18.0 · macOS 27.0 (26A428) · 2026-10-16mac-studio-m4-max gpu: pass stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.18.0 · macOS 27.0 (26A428) · 2026-10-16mac-studio-m4-max gpu: pass stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.18.0 · macOS 27.0 (26A428) · 2026-10-16mac-studio-m4-max gpu: pass stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.18.0 · macOS 27.0 (26A428) · 2026-10-16 | CARD.md · card.json | — |
qwen38-27b-mixed-int4 | text-generation | — | not measured | not measured | mac-studio-m4-max cpu: pass stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.18.1 · macOS 27.0 (26A428) · 2026-10-23mac-studio-m4-max gpu: pass stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.18.1 · macOS 27.0 (26A428) · 2026-10-23mac-studio-m4-max gpu: pass stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.18.1 · macOS 27.0 (26A428) · 2026-10-23mac-studio-m4-max gpu: pass stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.18.1 · macOS 27.0 (26A428) · 2026-10-23mac-studio-m4-max gpu: pass · 18.42 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.18.1 · macOS 27.0 (26A428) · 2026-10-23 | CARD.md · card.json | — |
r1-distill-qwen-1.5b | text-generation | — | not measured | not measured | mac-studio-m4-max gpu: pass stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.14.0 · 2026-07-22 | CARD.md · card.json | — |
ram-plus__ram_reweight_fp16 | image-classification | — | not measured | not measured | galaxy-s26 gpu_mldrift: load_failed stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 · 2026-08-26galaxy-s26 npu_qnn: pass · 28.303 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · QAIRT (Hexagon, JIT on-device) · Android 16 · 2026-08-26raspberry-pi-5 cpu_xnnpack: pass · 499.316 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | — |
ram-plus__ram_stage3_tail_fp16 | image-classification | — | not measured | not measured | galaxy-s26 gpu_mldrift: pass · 8.005 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 · 2026-08-26galaxy-s26 npu_qnn: pass · 14.568 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · QAIRT (Hexagon, JIT on-device) · Android 16 · 2026-08-26raspberry-pi-5 cpu_xnnpack: pass · 235.099 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | — |
ram-plus__ram_swin_s012_fp16 | image-classification | — | not measured | not measured | galaxy-s26 gpu_mldrift: pass · 102.984 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 · 2026-08-26galaxy-s26 npu_qnn: pass · 396.868 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · QAIRT (Hexagon, JIT on-device) · Android 16 · 2026-08-26raspberry-pi-5 cpu_xnnpack: pass · 3231.892 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | — |
ram-plus__ram_taghead_fp16 | image-classification | — | not measured | not measured | galaxy-s26 gpu_mldrift: pass · 44.03 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 · 2026-08-26galaxy-s26 npu_qnn: pass · 170.827 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · QAIRT (Hexagon, JIT on-device) · Android 16 · 2026-08-26raspberry-pi-5 cpu_xnnpack: pass · 1935.059 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | — |
rf-detr-nano__rfdetr_grapha_fp16 | object-detection | — | pass · 1686.515 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | pass · 18.487 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | galaxy-s26 gpu_mldrift: pass · 18.452 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 · 2026-08-26galaxy-s26 npu_qnn: pass · 15.11 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · QAIRT (Hexagon, JIT on-device) · Android 16 · 2026-08-26raspberry-pi-5 cpu_xnnpack: pass · 458.409 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
rf-detr-nano__rfdetr_graphb_fp16 | object-detection | — | pass · 134.482 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | output_mismatch · 4.797 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | galaxy-s26 gpu_mldrift: pass · 7.414 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 · 2026-08-25galaxy-s26 npu_qnn: pass · 12.822 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · QAIRT (Hexagon, JIT on-device) · Android 16 · 2026-08-25raspberry-pi-5 cpu_xnnpack: pass · 64.035 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
rf-detr-seg-nano__rfdetrseg_grapha_fp16 | image-segmentation | — | pass · 2194.563 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-20 | output_mismatch · 24.917 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-20 | galaxy-s26 gpu_mldrift: pass · 35.283 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 · 2026-08-26galaxy-s26 npu_qnn: pass · 27.833 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · QAIRT (Hexagon, JIT on-device) · Android 16 · 2026-08-26raspberry-pi-5 cpu_xnnpack: pass · 695.348 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
rf-detr-seg-nano__rfdetrseg_graphb_fp16 | image-segmentation | — | pass · 445.383 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-20 | output_mismatch · 16.68 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-20 | galaxy-s26 gpu_mldrift: pass · 22.663 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 · 2026-08-25galaxy-s26 npu_qnn: pass · 64.07 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · QAIRT (Hexagon, JIT on-device) · Android 16 · 2026-08-25raspberry-pi-5 cpu_xnnpack: pass · 324.54 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
rt-detrv2-s__rtdetr_grapha_fp16 | object-detection | — | pass · 3315.158 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | pass · 52.588 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | galaxy-s26 gpu_mldrift: pass · 23.066 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 · 2026-08-26galaxy-s26 npu_qnn: pass · 21.128 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · QAIRT (Hexagon, JIT on-device) · Android 16 · 2026-08-26raspberry-pi-5 cpu_xnnpack: pass · 523.099 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
rt-detrv2-s__rtdetr_graphb_fp16 | object-detection | — | pass · 634.398 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | pass · 27.723 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | galaxy-s26 gpu_mldrift: pass · 166.556 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 · 2026-08-25galaxy-s26 npu_qnn: pass · 132.551 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · QAIRT (Hexagon, JIT on-device) · Android 16 · 2026-08-25raspberry-pi-5 cpu_xnnpack: pass · 1199.458 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
rtmpose-animal-ap10k | keypoint-detection | — | pass · 335.895 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | pass · 6.217 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | galaxy-s26 gpu_mldrift: pass · 5.558 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 · 2026-08-26galaxy-s26 npu_qnn: pass · 1.289 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · QAIRT (Hexagon, JIT on-device) · Android 16 · 2026-08-26raspberry-pi-5 cpu_xnnpack: pass · 58.264 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
rtmpose-face-wflw | keypoint-detection | — | pass · 363.63 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | pass · 7.4 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | galaxy-s26 gpu_mldrift: pass · 6.769 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 · 2026-08-26galaxy-s26 npu_qnn: pass · 1.903 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · QAIRT (Hexagon, JIT on-device) · Android 16 · 2026-08-26raspberry-pi-5 cpu_xnnpack: pass · 71.582 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
rtmpose-hand | keypoint-detection | — | pass · 337.08 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | pass · 6.175 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | galaxy-s26 gpu_mldrift: pass · 5.715 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 · 2026-08-26galaxy-s26 npu_qnn: pass · 1.253 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · QAIRT (Hexagon, JIT on-device) · Android 16 · 2026-08-26raspberry-pi-5 cpu_xnnpack: pass · 58.101 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
rtmpose-s | keypoint-detection | — | pass · 90.525 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | pass · 2.96 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | galaxy-s26 gpu_mldrift: pass · 3.359 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 · 2026-08-25galaxy-s26 npu_qnn: pass · 0.771 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · QAIRT (Hexagon, JIT on-device) · Android 16 · 2026-08-25raspberry-pi-5 cpu_xnnpack: pass · 16.69 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
rtmw-m-wholebody | keypoint-detection | — | pass · 561.783 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | pass · 13.322 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | galaxy-s26 gpu_mldrift: pass · 11.882 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 · 2026-08-26galaxy-s26 npu_qnn: pass · 2.659 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · QAIRT (Hexagon, JIT on-device) · Android 16 · 2026-08-26raspberry-pi-5 cpu_xnnpack: pass · 109.617 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
rwkv-7-world-0.1b | text-generation | — | pass · 22.328 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | pass · 15.07 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | not measured | CARD.md · card.json | run it |
s1-mini-int8 | text-generation | — | not measured | not measured | galaxy-s26 cpu: fallback · 12.34 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-05galaxy-s26 gpu: pass · 25.8 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-05iphone-17-pro cpu: pass · 15.53 tok/s decode stale — latest known 0.19.1iphone-17-pro · A19-Pro · litert-lm 0.16.0 · iOS 27.0 · 2026-08-25iphone-17-pro gpu: pass · 32.02 tok/s decode stale — latest known 0.19.1iphone-17-pro · A19-Pro · litert-lm 0.16.0 · iOS 27.0 · 2026-08-25mac-studio-m4-max cpu: pass · 33.32 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · litert-lm 0.16.0 · 2026-08-25mac-studio-m4-max gpu: pass · 144.58 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · litert-lm 0.16.0 · 2026-08-25pixel-8a cpu: fallback · 5.77 tok/s decode stale — latest known 0.19.1Pixel 8a · Tensor G3 · litert-lm 0.16.0 · 2026-08-25pixel-8a gpu: pass · 12.28 tok/s decode stale — latest known 0.19.1Pixel 8a · Tensor G3 · litert-lm 0.16.0 · 2026-08-25 | CARD.md · card.json | — |
sam2.1-hiera-tiny-mask-decoder__sam2_tiny_mask_decoder_fp16 | mask-generation | — | pass · 396.69 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-13 | output_mismatch · 8.33 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-13 | galaxy-s26 gpu_mldrift: pass · 10.077 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 · 2026-08-26galaxy-s26 npu_qnn: pass · 8.08 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · QAIRT (Hexagon, JIT on-device) · Android 16 · 2026-08-26pixel-8a cpu_xnnpack: fallback · 218.668 ms stale — latest known 2.3.0Pixel 8a · Tensor G3 · litert 2.2.0 · Android 16 · 2026-08-28pixel-8a gpu_mldrift: pass · 30.055 ms stale — latest known 2.3.0Pixel 8a · Tensor G3 · litert 2.2.0 · Android 16 · 2026-08-28raspberry-pi-5 cpu_xnnpack: pass · 156.218 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
sam2.1-hiera-tiny-mask-decoder__sam2_tiny_mask_decoder_v2_fp16 | mask-generation | — | pass · 398.045 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-13 | output_mismatch · 7.908 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-13 | galaxy-s26 gpu_mldrift: pass · 10.548 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 · 2026-08-26galaxy-s26 npu_qnn: pass · 8.045 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · QAIRT (Hexagon, JIT on-device) · Android 16 · 2026-08-26pixel-8a cpu_xnnpack: fallback · 222.813 ms stale — latest known 2.3.0Pixel 8a · Tensor G3 · litert 2.2.0 · Android 16 · 2026-08-28pixel-8a gpu_mldrift: pass · 30.875 ms stale — latest known 2.3.0Pixel 8a · Tensor G3 · litert 2.2.0 · Android 16 · 2026-08-28raspberry-pi-5 cpu_xnnpack: pass · 158.27 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
sarashina2.2-0.5b-instruct-v0.1-int4 | text-generation | — | not measured | not measured | galaxy-s26 cpu: fallback · 19.53 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-02galaxy-s26 gpu: pass · 40.3 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-02mac-studio-m4-max cpu: pass · 42.73 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · litert-lm 0.16.0 · 2026-09-02mac-studio-m4-max gpu: pass · 197.85 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · litert-lm 0.16.0 · 2026-09-02 | CARD.md · card.json | — |
sarashina2.2-0.5b-instruct-v0.1-int8 | text-generation | — | not measured | not measured | galaxy-s26 cpu: fallback · 21.04 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-02galaxy-s26 gpu: pass · 36.46 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-02mac-studio-m4-max cpu: pass · 47.05 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · litert-lm 0.16.0 · 2026-09-02mac-studio-m4-max gpu: pass · 197.1 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · litert-lm 0.16.0 · 2026-09-02pixel-8a cpu: fallback · 8.78 tok/s decode stale — latest known 0.19.1Pixel 8a · Tensor G3 · litert-lm 0.16.0 · Android 16 · 2026-09-05pixel-8a gpu: pass · 16.79 tok/s decode stale — latest known 0.19.1Pixel 8a · Tensor G3 · litert-lm 0.16.0 · Android 16 · 2026-09-05 | CARD.md · card.json | — |
sarashina2.2-1b-instruct-v0.1-int4 | text-generation | — | not measured | not measured | galaxy-s26 cpu: fallback · 13.53 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-02galaxy-s26 gpu: pass · 27.45 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-02mac-studio-m4-max cpu: pass · 34.69 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · litert-lm 0.16.0 · 2026-09-02mac-studio-m4-max gpu: pass · 159.36 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · litert-lm 0.16.0 · 2026-09-02 | CARD.md · card.json | — |
sarashina2.2-1b-instruct-v0.1-int8 | text-generation | — | not measured | not measured | galaxy-s26 cpu: fallback · 13.35 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-02galaxy-s26 gpu: pass · 23.7 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-02mac-studio-m4-max cpu: pass · 35.14 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · litert-lm 0.16.0 · 2026-09-02mac-studio-m4-max gpu: pass · 146.4 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · litert-lm 0.16.0 · 2026-09-02pixel-8a cpu: fallback · 9.5 tok/s decode stale — latest known 0.19.1Pixel 8a · Tensor G3 · litert-lm 0.16.0 · Android 16 · 2026-09-07 | CARD.md · card.json | — |
shieldstral-1.0-3b-int8 | text-classification | — | not measured | not measured | galaxy-s26 cpu: fallback · 8.3 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-05galaxy-s26 gpu: pass · 8.63 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-08-24 | CARD.md · card.json | — |
shieldstral-1.0-3b-vision-int4 | image-text-to-text | — | not measured | not measured | galaxy-s26 cpu: fallback · 1.87 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-05galaxy-s26 gpu: load_failed stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-08-24 | CARD.md · card.json | — |
shieldstral-1.0-3b-vision-int4-gpu | image-text-to-text | — | not measured | not measured | galaxy-s26 cpu: fallback · 1.33 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-05galaxy-s26 gpu: pass · 11.73 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-08-28 | CARD.md · card.json | — |
siglip2-base-patch16-224 | image-feature-extraction | — | pass · 2124.707 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-13 | pass · 19.635 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-13 | galaxy-s26 gpu_mldrift: pass · 13.704 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 · 2026-08-26galaxy-s26 npu_qnn: pass · 6.943 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · QAIRT (Hexagon, JIT on-device) · Android 16 · 2026-08-26 | CARD.md · card.json | run it |
silent-face-anti-spoofing | image-classification | — | pass · 1.953 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | pass · 1.385 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | raspberry-pi-5 cpu_xnnpack: pass · 2.132 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
sinet-v2-camouflage | image-segmentation | — | pass · 349.405 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | output_mismatch · 13.072 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | raspberry-pi-5 cpu_xnnpack: pass · 274.669 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
smollm3-3b | text-generation | — | not measured | not measured | galaxy-s26 cpu: fallback · 14.24 tok/s decodeGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm prebuilt-adac974b · Android 16 · 2026-10-05galaxy-s26 cpu: fallback · 9.81 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-05galaxy-s26 cpu: fallback · 11.49 tok/s decodeGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm prebuilt-adac974b · Android 16 · 2026-10-05galaxy-s26 gpu: pass · 24.27 tok/s decodeGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm prebuilt-adac974b · Android 16 · 2026-10-05galaxy-s26 gpu: pass · 14.84 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-08-24galaxy-s26 gpu: pass · 24.36 tok/s decodeGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm prebuilt-adac974b · Android 16 · 2026-10-05iphone-18-pro cpu: fallback · 19.07 tok/s decode stale — latest known 0.19.1iPhone 18 Pro (iPhone19,2) · litert-lm 0.17.0 · iOS 27.0 · 2026-10-05iphone-18-pro cpu: fallback · 22.5 tok/s decode stale — latest known 0.19.1iPhone 18 Pro (iPhone19,2) · litert-lm 0.17.0 · iOS 27.0 · 2026-10-05iphone-18-pro gpu: pass · 43.33 tok/s decode stale — latest known 0.19.1iPhone 18 Pro (iPhone19,2) · litert-lm 0.17.0 · iOS 27.0 · 2026-10-05iphone-18-pro gpu: pass · 43.86 tok/s decode stale — latest known 0.19.1iPhone 18 Pro (iPhone19,2) · litert-lm 0.17.0 · iOS 27.0 · 2026-10-05mac-studio-m4-max cpu: fallback · 21.46 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.17.1 · 2026-10-05mac-studio-m4-max cpu: fallback · 24.53 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.17.1 · 2026-10-05mac-studio-m4-max gpu: pass · 124.74 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.17.1 · 2026-10-05mac-studio-m4-max gpu: pass · 131.82 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.17.1 · 2026-10-05pixel-8a gpu: pass · 7.69 tok/s decode stale — latest known 0.19.1Pixel 8a · Tensor G3 · litert-lm 0.16.0 · 2026-08-17raspberry-pi-5 cpu: pass · 2.46 tok/s decode stale — latest known 0.19.1Raspberry Pi 5 Model B Rev 1.1 · Broadcom BCM2712 · litert-lm 0.16.1 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-09-01 | CARD.md · card.json | — |
smollm3-3b-int8 | text-generation | — | not measured | not measured | galaxy-s26 cpu: fallback · 9.7 tok/s decodeGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm prebuilt-adac974b · Android 16 · 2026-10-05galaxy-s26 gpu: pass · 16.73 tok/s decodeGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm prebuilt-adac974b · Android 16 · 2026-10-05galaxy-s26 gpu: pass · 14.41 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-08-24galaxy-s26 gpu: pass · 16.02 tok/s decodeGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm prebuilt-adac974b · Android 16 · 2026-10-05galaxy-s26 npu: load_failedGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm selfbuilt-3dbb23e1e · QAIRT 2.47.0.260601114230 (QNN JIT libraries) · Android 16 · 2026-10-05mac-studio-m4-max cpu: fallback · 11.07 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.17.1 · 2026-10-05mac-studio-m4-max cpu: fallback · 24.18 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.17.1 · 2026-10-05mac-studio-m4-max gpu: pass · 100.22 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.17.1 · 2026-10-05mac-studio-m4-max gpu: pass · 105.96 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.17.1 · 2026-10-05raspberry-pi-5 cpu: pass · 1.97 tok/s decode stale — latest known 0.19.1Raspberry Pi 5 Model B Rev 1.1 · Broadcom BCM2712 · litert-lm 0.16.1 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-09-01 | CARD.md · card.json | — |
smollm3-3b-litert | text-generation | — | not measured | not measured | mac-studio-m4-max cpu: pass stale — latest known 0.19.1Mac Studio (M4 Max) · litert-lm 0.17.0 · 2026-09-06 | CARD.md · card.json | — |
smolvlm2-500m | image-text-to-text | — | not measured | not measured | galaxy-s26 cpu: pass · 72.41 tok/s decodeGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm selfbuilt-1dadd00c · Android 16 · 2026-09-26galaxy-s26 cpu: fallback · 76.72 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-05galaxy-s26 gpu: pass · 54.48 tok/s decodeGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm selfbuilt-1dadd00c · Android 16 · 2026-09-26galaxy-s26 gpu: pass · 76.73 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-08-25mac-studio-m4-max cpu: pass · 95.53 tok/s decodeMac Studio (M4 Max) · Apple M4 Max · litert-lm selfbuilt-1dadd00c · macOS 27.0 (26A428) · 2026-09-24mac-studio-m4-max gpu: run_failedMac Studio (M4 Max) · Apple M4 Max · litert-lm selfbuilt-1dadd00c · macOS 27.0 (26A428) · 2026-09-24 | CARD.md · card.json | — |
spark-x2.5-1.7b-int4 | text-generation | — | not measured | not measured | galaxy-s26 cpu: fallback · 13.6 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-07galaxy-s26 gpu: pass · 17.77 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-07iphone-17-pro cpu: pass stale — latest known 0.19.1iPhone 17 Pro · A19-Pro · litert-lm 0.15.0 · iOS 27.0 · 2026-09-07iphone-17-pro gpu: pass stale — latest known 0.19.1iPhone 17 Pro · A19-Pro · litert-lm 0.15.0 · iOS 27.0 · 2026-09-07mac-studio-m4-max cpu: pass · 38.24 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · litert-lm 0.17.0 · 2026-09-07mac-studio-m4-max gpu: pass · 103.62 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · litert-lm 0.17.0 · 2026-09-07 | CARD.md · card.json | — |
spark-x2.5-1.7b-int8 | text-generation | — | not measured | not measured | galaxy-s26 cpu: fallback · 12.28 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-07galaxy-s26 gpu: pass · 15.82 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-07iphone-17-pro cpu: pass stale — latest known 0.19.1iPhone 17 Pro · A19-Pro · litert-lm 0.15.0 · iOS 27.0 · 2026-09-07iphone-17-pro gpu: pass stale — latest known 0.19.1iPhone 17 Pro · A19-Pro · litert-lm 0.15.0 · iOS 27.0 · 2026-09-07mac-studio-m4-max cpu: pass · 41 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · litert-lm 0.17.0 · 2026-09-07mac-studio-m4-max gpu: pass · 95.78 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · litert-lm 0.17.0 · 2026-09-07 | CARD.md · card.json | — |
spark-x2.5-4b-int4 | text-generation | — | not measured | not measured | galaxy-s26 cpu: fallback · 5.47 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-07galaxy-s26 gpu: pass · 4.89 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-07iphone-17-pro cpu: pass stale — latest known 0.19.1iPhone 17 Pro · A19-Pro · litert-lm 0.15.0 · iOS 27.0 · 2026-09-07iphone-17-pro gpu: pass stale — latest known 0.19.1iPhone 17 Pro · A19-Pro · litert-lm 0.15.0 · iOS 27.0 · 2026-09-07mac-studio-m4-max cpu: pass · 18.37 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · litert-lm 0.17.0 · 2026-09-07mac-studio-m4-max gpu: pass · 53.63 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · litert-lm 0.17.0 · 2026-09-07 | CARD.md · card.json | — |
spark-x2.5-4b-int8 | text-generation | — | not measured | not measured | galaxy-s26 cpu: fallback · 4.55 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-07galaxy-s26 gpu: pass stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-07iphone-17-pro cpu: pass stale — latest known 0.19.1iPhone 17 Pro · A19-Pro · litert-lm 0.15.0 · iOS 27.0 · 2026-09-07iphone-17-pro gpu: pass stale — latest known 0.19.1iPhone 17 Pro · A19-Pro · litert-lm 0.15.0 · iOS 27.0 · 2026-09-07mac-studio-m4-max cpu: pass · 19.5 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · litert-lm 0.17.0 · 2026-09-07mac-studio-m4-max gpu: pass · 49.38 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · litert-lm 0.17.0 · 2026-09-07 | CARD.md · card.json | — |
speaker-diarization | automatic-speech-recognition# HF card front matter pipeline_tag | — | pass · 1399.142 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | pass · 3.852 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | galaxy-s26 gpu_mldrift: pass · 6.05 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 · 2026-08-25galaxy-s26 npu_qnn: pass · 1.993 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · QAIRT (Hexagon, JIT on-device) · Android 16 · 2026-08-25raspberry-pi-5 cpu_xnnpack: pass · 186.062 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
tashkeel-350m-v2-fp16 | text-generation | — | not measured | not measured | galaxy-s26 cpu: fallback · 23.69 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-05mac-studio-m4-max cpu: pass · 58.18 tok/s decode stale — latest known 0.19.1mac-studio-m4-max · Apple M4 Max · litert-lm 0.16.0 · 2026-08-25pixel-8a cpu: fallback · 14.42 tok/s decode stale — latest known 0.19.1Pixel 8a · Tensor G3 · litert-lm 0.16.0 · Android 16 · 2026-09-05 | CARD.md · card.json | — |
tashkeel-350m-v2-int8 | text-generation | — | not measured | not measured | galaxy-s26 cpu: fallback · 68.96 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-05galaxy-s26 gpu: pass · 46.26 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-05mac-studio-m4-max cpu: pass · 97.21 tok/s decode stale — latest known 0.19.1mac-studio-m4-max · Apple M4 Max · litert-lm 0.16.0 · 2026-08-25pixel-8a cpu: fallback · 49.06 tok/s decode stale — latest known 0.19.1Pixel 8a · Tensor G3 · litert-lm 0.16.0 · Android 16 · 2026-09-05pixel-8a gpu: pass · 18.37 tok/s decode stale — latest known 0.19.1Pixel 8a · Tensor G3 · litert-lm 0.16.0 · Android 16 · 2026-09-05 | CARD.md · card.json | — |
tiger-dnr__tiger_dialog_fp16 | audio-to-audio | — | pass · 27817.165 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | output_mismatch · 483.59 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | galaxy-s26 npu_qnn: load_failed stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · QAIRT (Hexagon, JIT on-device) · Android 16 · 2026-08-26raspberry-pi-5 cpu_xnnpack: pass · 42421.191 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
tiger-dnr__tiger_effect_fp16 | audio-to-audio | — | pass · 27916.245 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | output_mismatch · 498.37 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | galaxy-s26 npu_qnn: load_failed stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · QAIRT (Hexagon, JIT on-device) · Android 16 · 2026-08-26raspberry-pi-5 cpu_xnnpack: pass · 42412.096 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
tiger-dnr__tiger_music_fp16 | audio-to-audio | — | pass · 28192.605 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | output_mismatch · 497.43 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | galaxy-s26 npu_qnn: load_failed stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · QAIRT (Hexagon, JIT on-device) · Android 16 · 2026-08-26raspberry-pi-5 cpu_xnnpack: pass · 42479.118 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
timestral-209m__timestral-209m-c2048-h1024 | time-series-forecasting | gpu_mldrift @ litert 2.2.0: 85.4% · 276 partition(s) | not measured | not measured | galaxy-s26 cpu_xnnpack: pass · 93.7570575 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU, Hexagon HTP) · litert 2.2.0 · Android 16 BP4A.251205.006.S942QOPS1AZH9 · 2026-10-22galaxy-s26 gpu_mldrift: pass · 28.838151 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU, Hexagon HTP) · litert 2.2.0 · Android 16 BP4A.251205.006.S942QOPS1AZH9 · 2026-10-22 | CARD.md · card.json | — |
timestral-209m__timestral-209m-c2048-h1024-int8 | time-series-forecasting | gpu_mldrift @ litert 2.2.0: 95.5% · 77 partition(s) | not measured | not measured | galaxy-s26 cpu_xnnpack: pass · 39.298698 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU, Hexagon HTP) · litert 2.2.0 · Android 16 BP4A.251205.006.S942QOPS1AZH9 · 2026-10-22 | CARD.md · card.json | — |
timestral-209m__timestral-209m-c2048-h256 | time-series-forecasting | gpu_mldrift @ litert 2.2.0: 85.4% · 276 partition(s) | not measured | not measured | galaxy-s26 cpu_xnnpack: pass · 76.1672915 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU, Hexagon HTP) · litert 2.2.0 · Android 16 BP4A.251205.006.S942QOPS1AZH9 · 2026-10-22galaxy-s26 gpu_mldrift: pass · 27.0026565 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU, Hexagon HTP) · litert 2.2.0 · Android 16 BP4A.251205.006.S942QOPS1AZH9 · 2026-10-22galaxy-s26 npu_qnn: pass · 15.569500000000001 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU, Hexagon HTP) · litert 2.2.0 · Qualcomm QAIRT QNN HTP libraries (libQnnHtp.so v2.47.0.260601114230, v81 skel) + the LiteRT 2.2.0 release's Qualcomm dispatch library and JIT compiler plugin · Android 16 BP4A.251205.006.S942QOPS1AZH9 · 2026-10-22 | CARD.md · card.json | — |
timestral-209m__timestral-209m-c2048-h256-int8 | time-series-forecasting | gpu_mldrift @ litert 2.2.0: 95.5% · 77 partition(s) | not measured | not measured | galaxy-s26 cpu_xnnpack: pass · 30.252813 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU, Hexagon HTP) · litert 2.2.0 · Android 16 BP4A.251205.006.S942QOPS1AZH9 · 2026-10-22 | CARD.md · card.json | — |
timestral-209m__timestral-209m-c8192-h1024 | time-series-forecasting | gpu_mldrift @ litert 2.2.0: 85.4% · 276 partition(s) | not measured | not measured | galaxy-s26 cpu_xnnpack: pass · 321.000547 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU, Hexagon HTP) · litert 2.2.0 · Android 16 BP4A.251205.006.S942QOPS1AZH9 · 2026-10-22galaxy-s26 gpu_mldrift: pass · 73.4220055 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU, Hexagon HTP) · litert 2.2.0 · Android 16 BP4A.251205.006.S942QOPS1AZH9 · 2026-10-22 | CARD.md · card.json | — |
timestral-209m__timestral-209m-c8192-h1024-int8 | time-series-forecasting | gpu_mldrift @ litert 2.2.0: 95.5% · 77 partition(s) | not measured | not measured | galaxy-s26 cpu_xnnpack: pass · 144.1709895 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU, Hexagon HTP) · litert 2.2.0 · Android 16 BP4A.251205.006.S942QOPS1AZH9 · 2026-10-22 | CARD.md · card.json | — |
timestral-209m__timestral-209m-c8192-h256 | time-series-forecasting | gpu_mldrift @ litert 2.2.0: 85.4% · 276 partition(s) | not measured | not measured | galaxy-s26 cpu_xnnpack: pass · 294.1965365 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU, Hexagon HTP) · litert 2.2.0 · Android 16 BP4A.251205.006.S942QOPS1AZH9 · 2026-10-22galaxy-s26 gpu_mldrift: pass · 68.714922 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU, Hexagon HTP) · litert 2.2.0 · Android 16 BP4A.251205.006.S942QOPS1AZH9 · 2026-10-22 | CARD.md · card.json | — |
timestral-209m__timestral-209m-c8192-h256-int8 | time-series-forecasting | gpu_mldrift @ litert 2.2.0: 95.5% · 77 partition(s) | not measured | not measured | galaxy-s26 cpu_xnnpack: pass · 128.49638 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU, Hexagon HTP) · litert 2.2.0 · Android 16 BP4A.251205.006.S942QOPS1AZH9 · 2026-10-22 | CARD.md · card.json | — |
tipsv2-b14-dpt | depth-estimation | — | pass · 148.665 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-20 | output_mismatch · 285.11 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-20 | not measured | CARD.md · card.json | run it |
twinlitenet | image-segmentation | — | pass · 223.485 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | output_mismatch · 19.31 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | raspberry-pi-5 cpu_xnnpack: pass · 261.197 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
u2net-portrait-sketch | image-to-image | — | pass · 4289.642 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | pass · 53.318 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | raspberry-pi-5 cpu_xnnpack: pass · 2871.872 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
ultra-fast-lane-detection | image-segmentation | — | pass · 238.387 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | pass · 7.463 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | raspberry-pi-5 cpu_xnnpack: pass · 150.799 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
unisal-saliency | image-segmentation | — | pass · 217.372 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | pass · 3.402 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | galaxy-s26 gpu_mldrift: pass · 3.658 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 · 2026-08-25galaxy-s26 npu_qnn: pass · 21.121 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · QAIRT (Hexagon, JIT on-device) · Android 16 · 2026-08-25raspberry-pi-5 cpu_xnnpack: pass · 102.024 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
vela-1.0-encoder-307m-embedding__vela_l128_embeds_fp16 | sentence-similarity | gpu_mldrift @ litert 2.2.0: 92.1% · 134 partition(s) | pass · 399.72 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-10-18 | pass · 19.13 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-10-18 | galaxy-s26 gpu_mldrift: pass stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU, Hexagon HTP) · litert 2.2.0 · Android SDK 36 (Build.VERSION.SDK_INT read by the gate app; Build.MODEL SM-S942Q) · 2026-10-31galaxy-s26 gpu_mldrift: output_mismatch · 49.399167 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU, Hexagon HTP) · litert 2.2.0 · Android SDK 36 (Build.VERSION.SDK_INT read by the gate app; Build.MODEL SM-S942Q) · 2026-10-18galaxy-s26 gpu_mldrift: pass stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU, Hexagon HTP) · litert 2.2.0 · Android SDK 36 (Build.VERSION.SDK_INT read by the gate app; Build.MODEL SM-S942Q) · 2026-10-31galaxy-s26 gpu_mldrift: pass · 24.098672 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU, Hexagon HTP) · litert 2.2.0 · Android SDK 36 (Build.VERSION.SDK_INT read by the gate app; Build.MODEL SM-S942Q) · 2026-10-18galaxy-s26 npu_qnn: pass · 13.062 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU, Hexagon HTP) · litert 2.2.0 · Qualcomm QNN HTP libraries (v81) in /data/local/tmp/vela_npu · Android SDK 36 (Build.VERSION.SDK_INT read by the gate app; Build.MODEL SM-S942Q) · 2026-10-18pixel-8a gpu_mldrift: pass stale — latest known 2.3.0Pixel 8a · Google Tensor G3 (Mali-G715 GPU) · litert 2.2.0 · Android SDK 36 (Build.VERSION.SDK_INT read by the gate app; Build.MODEL Pixel 8a) · 2026-10-31pixel-8a gpu_mldrift: output_mismatch · 67.639405 ms stale — latest known 2.3.0Pixel 8a · Google Tensor G3 (Mali-G715 GPU) · litert 2.2.0 · Android SDK 36 (Build.VERSION.SDK_INT read by the gate app; Build.MODEL Pixel 8a) · 2026-10-18pixel-8a gpu_mldrift: output_mismatch · 71.886922 ms stale — latest known 2.3.0Pixel 8a · Google Tensor G3 (Mali-G715 GPU) · litert 2.2.0 · Android SDK 36 (Build.VERSION.SDK_INT read by the gate app; Build.MODEL Pixel 8a) · 2026-10-18pixel-8a gpu_mldrift: pass stale — latest known 2.3.0Pixel 8a · Google Tensor G3 (Mali-G715 GPU) · litert 2.2.0 · Android SDK 36 (Build.VERSION.SDK_INT read by the gate app; Build.MODEL Pixel 8a) · 2026-10-31pixel-8a gpu_mldrift: pass · 91.464844 ms stale — latest known 2.3.0Pixel 8a · Google Tensor G3 (Mali-G715 GPU) · litert 2.2.0 · Android SDK 36 (Build.VERSION.SDK_INT read by the gate app; Build.MODEL Pixel 8a) · 2026-10-18 | CARD.md · card.json | run it |
vela-1.0-encoder-307m-embedding__vela_l128_fp16 | sentence-similarity | gpu_mldrift @ litert 2.2.0: 92% · 134 partition(s) | not measured | not measured | galaxy-s26 cpu_xnnpack: fallback · 41.3828645 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU, Hexagon HTP) · litert 2.2.0 · Android SDK 36 (Build.VERSION.SDK_INT read by the gate app; Build.MODEL SM-S942Q) · 2026-10-18galaxy-s26 npu_qnn: pass · 13.076 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU, Hexagon HTP) · litert 2.2.0 · Qualcomm QNN HTP libraries (v81) in /data/local/tmp/vela_npu · Android SDK 36 (Build.VERSION.SDK_INT read by the gate app; Build.MODEL SM-S942Q) · 2026-10-18pixel-8a cpu_xnnpack: fallback · 472.694051 ms stale — latest known 2.3.0Pixel 8a · Google Tensor G3 (Mali-G715 GPU) · litert 2.2.0 · Android SDK 36 (Build.VERSION.SDK_INT read by the gate app; Build.MODEL Pixel 8a) · 2026-10-18 | CARD.md · card.json | — |
vela-1.0-encoder-307m-embedding__vela_l256_embeds_fp16 | sentence-similarity | gpu_mldrift @ litert 2.2.0: 92.1% · 134 partition(s) | pass · 826.035 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-10-18 | pass · 22.592 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-10-18 | galaxy-s26 gpu_mldrift: pass stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU, Hexagon HTP) · litert 2.2.0 · Android SDK 36 (Build.VERSION.SDK_INT read by the gate app; Build.MODEL SM-S942Q) · 2026-10-31galaxy-s26 gpu_mldrift: output_mismatch · 93.34460949999999 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU, Hexagon HTP) · litert 2.2.0 · Android SDK 36 (Build.VERSION.SDK_INT read by the gate app; Build.MODEL SM-S942Q) · 2026-10-18galaxy-s26 gpu_mldrift: pass stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU, Hexagon HTP) · litert 2.2.0 · Android SDK 36 (Build.VERSION.SDK_INT read by the gate app; Build.MODEL SM-S942Q) · 2026-10-31galaxy-s26 gpu_mldrift: pass · 46.42276 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU, Hexagon HTP) · litert 2.2.0 · Android SDK 36 (Build.VERSION.SDK_INT read by the gate app; Build.MODEL SM-S942Q) · 2026-10-18galaxy-s26 npu_qnn: pass · 28.015 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU, Hexagon HTP) · litert 2.2.0 · Qualcomm QNN HTP libraries (v81) in /data/local/tmp/vela_npu · Android SDK 36 (Build.VERSION.SDK_INT read by the gate app; Build.MODEL SM-S942Q) · 2026-10-18pixel-8a gpu_mldrift: pass stale — latest known 2.3.0Pixel 8a · Google Tensor G3 (Mali-G715 GPU) · litert 2.2.0 · Android SDK 36 (Build.VERSION.SDK_INT read by the gate app; Build.MODEL Pixel 8a) · 2026-10-31pixel-8a gpu_mldrift: output_mismatch · 103.331706 ms stale — latest known 2.3.0Pixel 8a · Google Tensor G3 (Mali-G715 GPU) · litert 2.2.0 · Android SDK 36 (Build.VERSION.SDK_INT read by the gate app; Build.MODEL Pixel 8a) · 2026-10-18pixel-8a gpu_mldrift: output_mismatch · 113.181152 ms stale — latest known 2.3.0Pixel 8a · Google Tensor G3 (Mali-G715 GPU) · litert 2.2.0 · Android SDK 36 (Build.VERSION.SDK_INT read by the gate app; Build.MODEL Pixel 8a) · 2026-10-18pixel-8a gpu_mldrift: pass stale — latest known 2.3.0Pixel 8a · Google Tensor G3 (Mali-G715 GPU) · litert 2.2.0 · Android SDK 36 (Build.VERSION.SDK_INT read by the gate app; Build.MODEL Pixel 8a) · 2026-10-31pixel-8a gpu_mldrift: pass · 180.58221450000002 ms stale — latest known 2.3.0Pixel 8a · Google Tensor G3 (Mali-G715 GPU) · litert 2.2.0 · Android SDK 36 (Build.VERSION.SDK_INT read by the gate app; Build.MODEL Pixel 8a) · 2026-10-18 | CARD.md · card.json | run it |
vela-1.0-encoder-307m-embedding__vela_l256_fp16 | sentence-similarity | gpu_mldrift @ litert 2.2.0: 92% · 134 partition(s) | not measured | not measured | galaxy-s26 cpu_xnnpack: fallback · 88.03138 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU, Hexagon HTP) · litert 2.2.0 · Android SDK 36 (Build.VERSION.SDK_INT read by the gate app; Build.MODEL SM-S942Q) · 2026-10-18galaxy-s26 npu_qnn: pass · 29.049 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU, Hexagon HTP) · litert 2.2.0 · Qualcomm QNN HTP libraries (v81) in /data/local/tmp/vela_npu · Android SDK 36 (Build.VERSION.SDK_INT read by the gate app; Build.MODEL SM-S942Q) · 2026-10-18pixel-8a cpu_xnnpack: fallback · 935.7937830000001 ms stale — latest known 2.3.0Pixel 8a · Google Tensor G3 (Mali-G715 GPU) · litert 2.2.0 · Android SDK 36 (Build.VERSION.SDK_INT read by the gate app; Build.MODEL Pixel 8a) · 2026-10-18 | CARD.md · card.json | — |
vela-1.0-encoder-307m-embedding__vela_l512_embeds_fp16 | sentence-similarity | gpu_mldrift @ litert 2.2.0: 92.1% · 134 partition(s) | pass · 1789.8 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-10-18 | pass · 38.43 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-10-18 | galaxy-s26 gpu_mldrift: pass · 133.26350000000002 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU, Hexagon HTP) · litert 2.2.0 · Android SDK 36 (Build.VERSION.SDK_INT read by the gate app; Build.MODEL SM-S942Q) · 2026-10-18galaxy-s26 gpu_mldrift: pass stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU, Hexagon HTP) · litert 2.2.0 · Android SDK 36 (Build.VERSION.SDK_INT read by the gate app; Build.MODEL SM-S942Q) · 2026-10-31galaxy-s26 gpu_mldrift: output_mismatch · 100.54872399999999 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU, Hexagon HTP) · litert 2.2.0 · Android SDK 36 (Build.VERSION.SDK_INT read by the gate app; Build.MODEL SM-S942Q) · 2026-10-18galaxy-s26 gpu_mldrift: output_mismatch · 83.86 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU, Hexagon HTP) · litert 2.2.0 · Android SDK 36 (Build.VERSION.SDK_INT read by the gate app; Build.MODEL SM-S942Q) · 2026-10-18galaxy-s26 gpu_mldrift: pass stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU, Hexagon HTP) · litert 2.2.0 · Android SDK 36 (Build.VERSION.SDK_INT read by the gate app; Build.MODEL SM-S942Q) · 2026-10-31galaxy-s26 gpu_mldrift: pass · 119.5477865 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU, Hexagon HTP) · litert 2.2.0 · Android SDK 36 (Build.VERSION.SDK_INT read by the gate app; Build.MODEL SM-S942Q) · 2026-10-18galaxy-s26 npu_qnn: pass · 68.4005 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU, Hexagon HTP) · litert 2.2.0 · Qualcomm QNN HTP libraries (v81) in /data/local/tmp/vela_npu · Android SDK 36 (Build.VERSION.SDK_INT read by the gate app; Build.MODEL SM-S942Q) · 2026-10-18pixel-8a gpu_mldrift: pass stale — latest known 2.3.0Pixel 8a · Google Tensor G3 (Mali-G715 GPU) · litert 2.2.0 · Android SDK 36 (Build.VERSION.SDK_INT read by the gate app; Build.MODEL Pixel 8a) · 2026-10-31pixel-8a gpu_mldrift: output_mismatch · 209.0309655 ms stale — latest known 2.3.0Pixel 8a · Google Tensor G3 (Mali-G715 GPU) · litert 2.2.0 · Android SDK 36 (Build.VERSION.SDK_INT read by the gate app; Build.MODEL Pixel 8a) · 2026-10-18pixel-8a gpu_mldrift: output_mismatch · 230.8465575 ms stale — latest known 2.3.0Pixel 8a · Google Tensor G3 (Mali-G715 GPU) · litert 2.2.0 · Android SDK 36 (Build.VERSION.SDK_INT read by the gate app; Build.MODEL Pixel 8a) · 2026-10-18pixel-8a gpu_mldrift: pass stale — latest known 2.3.0Pixel 8a · Google Tensor G3 (Mali-G715 GPU) · litert 2.2.0 · Android SDK 36 (Build.VERSION.SDK_INT read by the gate app; Build.MODEL Pixel 8a) · 2026-10-31pixel-8a gpu_mldrift: pass · 408.3141075 ms stale — latest known 2.3.0Pixel 8a · Google Tensor G3 (Mali-G715 GPU) · litert 2.2.0 · Android SDK 36 (Build.VERSION.SDK_INT read by the gate app; Build.MODEL Pixel 8a) · 2026-10-18 | CARD.md · card.json | run it |
vela-1.0-encoder-307m-embedding__vela_l512_fp16 | sentence-similarity | gpu_mldrift @ litert 2.2.0: 92% · 134 partition(s) | not measured | not measured | galaxy-s26 cpu_xnnpack: fallback · 237.2797655 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU, Hexagon HTP) · litert 2.2.0 · Android SDK 36 (Build.VERSION.SDK_INT read by the gate app; Build.MODEL SM-S942Q) · 2026-10-18galaxy-s26 npu_qnn: pass · 71.033 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU, Hexagon HTP) · litert 2.2.0 · Qualcomm QNN HTP libraries (v81) in /data/local/tmp/vela_npu · Android SDK 36 (Build.VERSION.SDK_INT read by the gate app; Build.MODEL SM-S942Q) · 2026-10-18pixel-8a cpu_xnnpack: fallback · 2044.822877 ms stale — latest known 2.3.0Pixel 8a · Google Tensor G3 (Mali-G715 GPU) · litert 2.2.0 · Android SDK 36 (Build.VERSION.SDK_INT read by the gate app; Build.MODEL Pixel 8a) · 2026-10-18 | CARD.md · card.json | — |
vela-1.0-encoder-307m-embedding__vela_l64_embeds_fp16 | sentence-similarity | gpu_mldrift @ litert 2.2.0: 92.1% · 134 partition(s) | pass · 196.29 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-10-18 | pass · 14.09 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-10-18 | galaxy-s26 gpu_mldrift: pass · 11.5015 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU, Hexagon HTP) · litert 2.2.0 · Android SDK 36 (Build.VERSION.SDK_INT read by the gate app; Build.MODEL SM-S942Q) · 2026-10-18galaxy-s26 gpu_mldrift: pass · 17.0155 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU, Hexagon HTP) · litert 2.2.0 · Android SDK 36 (Build.VERSION.SDK_INT read by the gate app; Build.MODEL SM-S942Q) · 2026-10-18galaxy-s26 gpu_mldrift: pass stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU, Hexagon HTP) · litert 2.2.0 · Android SDK 36 (Build.VERSION.SDK_INT read by the gate app; Build.MODEL SM-S942Q) · 2026-10-31galaxy-s26 gpu_mldrift: output_mismatch · 11.201354 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU, Hexagon HTP) · litert 2.2.0 · Android SDK 36 (Build.VERSION.SDK_INT read by the gate app; Build.MODEL SM-S942Q) · 2026-10-18galaxy-s26 gpu_mldrift: output_mismatch · 28.803958 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU, Hexagon HTP) · litert 2.2.0 · Android SDK 36 (Build.VERSION.SDK_INT read by the gate app; Build.MODEL SM-S942Q) · 2026-10-18galaxy-s26 gpu_mldrift: pass stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU, Hexagon HTP) · litert 2.2.0 · Android SDK 36 (Build.VERSION.SDK_INT read by the gate app; Build.MODEL SM-S942Q) · 2026-10-31galaxy-s26 gpu_mldrift: pass · 16.482760499999998 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU, Hexagon HTP) · litert 2.2.0 · Android SDK 36 (Build.VERSION.SDK_INT read by the gate app; Build.MODEL SM-S942Q) · 2026-10-18galaxy-s26 npu_qnn: pass · 6.833 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU, Hexagon HTP) · litert 2.2.0 · Qualcomm QNN HTP libraries (v81) in /data/local/tmp/vela_npu · Android SDK 36 (Build.VERSION.SDK_INT read by the gate app; Build.MODEL SM-S942Q) · 2026-10-18pixel-8a gpu_mldrift: pass stale — latest known 2.3.0Pixel 8a · Google Tensor G3 (Mali-G715 GPU) · litert 2.2.0 · Android SDK 36 (Build.VERSION.SDK_INT read by the gate app; Build.MODEL Pixel 8a) · 2026-10-31pixel-8a gpu_mldrift: output_mismatch · 54.605306 ms stale — latest known 2.3.0Pixel 8a · Google Tensor G3 (Mali-G715 GPU) · litert 2.2.0 · Android SDK 36 (Build.VERSION.SDK_INT read by the gate app; Build.MODEL Pixel 8a) · 2026-10-18pixel-8a gpu_mldrift: output_mismatch · 55.912923 ms stale — latest known 2.3.0Pixel 8a · Google Tensor G3 (Mali-G715 GPU) · litert 2.2.0 · Android SDK 36 (Build.VERSION.SDK_INT read by the gate app; Build.MODEL Pixel 8a) · 2026-10-18pixel-8a gpu_mldrift: pass stale — latest known 2.3.0Pixel 8a · Google Tensor G3 (Mali-G715 GPU) · litert 2.2.0 · Android SDK 36 (Build.VERSION.SDK_INT read by the gate app; Build.MODEL Pixel 8a) · 2026-10-31pixel-8a gpu_mldrift: pass · 75.922119 ms stale — latest known 2.3.0Pixel 8a · Google Tensor G3 (Mali-G715 GPU) · litert 2.2.0 · Android SDK 36 (Build.VERSION.SDK_INT read by the gate app; Build.MODEL Pixel 8a) · 2026-10-18 | CARD.md · card.json | run it |
vela-1.0-encoder-307m-embedding__vela_l64_fp16 | sentence-similarity | gpu_mldrift @ litert 2.2.0: 91.9% · 134 partition(s) | not measured | not measured | galaxy-s26 cpu_xnnpack: fallback · 19.8360675 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU, Hexagon HTP) · litert 2.2.0 · Android SDK 36 (Build.VERSION.SDK_INT read by the gate app; Build.MODEL SM-S942Q) · 2026-10-18galaxy-s26 npu_qnn: pass · 6.823 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU, Hexagon HTP) · litert 2.2.0 · Qualcomm QNN HTP libraries (v81) in /data/local/tmp/vela_npu · Android SDK 36 (Build.VERSION.SDK_INT read by the gate app; Build.MODEL SM-S942Q) · 2026-10-18pixel-8a cpu_xnnpack: fallback · 180.72762 ms stale — latest known 2.3.0Pixel 8a · Google Tensor G3 (Mali-G715 GPU) · litert 2.2.0 · Android SDK 36 (Build.VERSION.SDK_INT read by the gate app; Build.MODEL Pixel 8a) · 2026-10-18pixel-8a gpu_mldrift: run_failed stale — latest known 2.3.0Pixel 8a · Google Tensor G3 (Mali-G715 GPU) · litert 2.2.0 · Android SDK 36 (Build.VERSION.SDK_INT read by the gate app; Build.MODEL Pixel 8a) · 2026-10-18 | CARD.md · card.json | — |
vibethinker-3b | text-generation | — | not measured | not measured | galaxy-s26 cpu: fallback · 16.09 tok/s decodeGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm prebuilt-adac974b · Android 16 · 2026-10-05galaxy-s26 gpu: pass · 23.54 tok/s decodeGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm prebuilt-adac974b · Android 16 · 2026-10-05galaxy-s26 gpu: pass · 13.88 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-08-24galaxy-s26 gpu: pass · 25.23 tok/s decodeGalaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm prebuilt-adac974b · Android 16 · 2026-10-05mac-studio-m4-max cpu: fallback · 24.73 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.17.1 · 2026-10-05mac-studio-m4-max cpu: fallback · 29.84 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.17.1 · 2026-10-05mac-studio-m4-max gpu: pass · 117.04 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.17.1 · 2026-10-05mac-studio-m4-max gpu: pass · 125.62 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · Apple M4 Max · litert-lm 0.17.1 · 2026-10-05pixel-8a gpu: pass · 7.58 tok/s decode stale — latest known 0.19.1Pixel 8a · Tensor G3 · litert-lm 0.16.0 · 2026-08-17raspberry-pi-5 cpu: pass · 3.36 tok/s decode stale — latest known 0.19.1Raspberry Pi 5 Model B Rev 1.1 · Broadcom BCM2712 · litert-lm 0.16.1 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-09-01 | CARD.md · card.json | — |
vibevoice-asr-bitnet | automatic-speech-recognition | — | not measured | not measured | galaxy-s26 cpu: fallback · 35.43 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.1 · Android 16 · 2026-09-04galaxy-s26 gpu: pass · 36.12 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.1 · Android 16 · 2026-09-04mac-studio-m4-max cpu: pass · 59.29 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · litert-lm 0.16.0 · 2026-09-04mac-studio-m4-max gpu: pass · 138.57 tok/s decode stale — latest known 0.19.1Mac Studio (M4 Max) · litert-lm 0.16.0 · 2026-09-04 | CARD.md · card.json | — |
vision-rwkv-s | image-classification | — | pass · 316.283 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | output_mismatch · 27.517 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | raspberry-pi-5 cpu_xnnpack: pass · 1082.624 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
wav2vec2-base-960h__w2v2_asr_frontend_fp16 | automatic-speech-recognition | — | run_failed stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-20 | pass · 37.113 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-20 | raspberry-pi-5 cpu_xnnpack: pass · 2010.862 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
wav2vec2-base-960h__w2v2_asr_head_fp16 | automatic-speech-recognition | — | run_failed stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-20 | pass · 51.323 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-20 | raspberry-pi-5 cpu_xnnpack: pass · 2820.028 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
wav2vec2-keyword-spotting__w2v2_frontend_fp16 | audio-classification | — | pass · 309.91 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | pass · 3.365 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | galaxy-s26 gpu_mldrift: pass · 5.285 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 · 2026-08-25galaxy-s26 npu_qnn: run_failed stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · QAIRT (Hexagon, JIT on-device) · Android 16 · 2026-08-25raspberry-pi-5 cpu_xnnpack: pass · 106.245 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
wav2vec2-keyword-spotting__w2v2_head_fp16 | audio-classification | — | pass · 685.443 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | pass · 12.15 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | galaxy-s26 gpu_mldrift: pass · 9.197 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 · 2026-08-26galaxy-s26 npu_qnn: pass · 5.668 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · QAIRT (Hexagon, JIT on-device) · Android 16 · 2026-08-26raspberry-pi-5 cpu_xnnpack: pass · 106.275 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
wevisone-vl-4b-int4 | image-text-to-text | — | not measured | not measured | galaxy-s26 gpu: passGalaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU) · litert-lm 0.19.1 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-10galaxy-s26 gpu: pass · 15.2 tok/s decodeGalaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU) · litert-lm 0.19.1 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-10galaxy-s26 gpu: pass · 15.23 tok/s decodeGalaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU) · litert-lm 0.19.1 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-10mac-studio-m4-max cpu: pass · 23.68 tok/s decodeMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.1 · macOS 27.0 (26A428) · 2026-11-12mac-studio-m4-max cpu: pass · 25.93 tok/s decodeMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.1 · macOS 27.0 (26A428) · 2026-11-12mac-studio-m4-max gpu: pass stale — latest known 0.19.1Mac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.18.1 · macOS 27.0 (26A428) · 2026-10-27mac-studio-m4-max gpu: pass · 92.34 tok/s decodeMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.1 · macOS 27.0 (26A428) · 2026-11-12mac-studio-m4-max gpu: pass · 89.17 tok/s decodeMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.1 · macOS 27.0 (26A428) · 2026-11-12 | CARD.md · card.json | — |
wevisone-vl-4b-int8 | image-text-to-text | — | not measured | not measured | galaxy-s26 cpu: passGalaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU) · litert-lm 0.19.1 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-10galaxy-s26 cpu: pass · 10.25 tok/s decodeGalaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU) · litert-lm 0.19.1 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-10galaxy-s26 cpu: pass · 8.24 tok/s decodeGalaxy S26 (SM-S942Q) · Qualcomm SM8850 (Adreno GPU) · litert-lm 0.19.1 · Android 16 (BP4A.251205.006.S942QOPS1AZH9) · 2026-11-10mac-studio-m4-max cpu: pass · 24.79 tok/s decodeMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.1 · macOS 27.0 (26A428) · 2026-11-12mac-studio-m4-max cpu: pass · 23.92 tok/s decodeMac Studio (M4 Max, Mac16,9) · Apple M4 Max · litert-lm 0.19.1 · macOS 27.0 (26A428) · 2026-11-12 | CARD.md · card.json | — |
xfeat__xfeat | keypoint-detection | — | pass · 189.138 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | pass · 12.725 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | galaxy-s26 gpu_mldrift: pass · 4.128 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 · 2026-08-25galaxy-s26 npu_qnn: pass · 607.499 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · QAIRT (Hexagon, JIT on-device) · Android 16 · 2026-08-25raspberry-pi-5 cpu_xnnpack: pass · 32.204 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
xfeat__xfeat_fp16 | keypoint-detection | — | pass · 189.197 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | pass · 12.378 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | galaxy-s26 gpu_mldrift: pass · 4.313 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 · 2026-08-25galaxy-s26 npu_qnn: pass · 384.541 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · QAIRT (Hexagon, JIT on-device) · Android 16 · 2026-08-25raspberry-pi-5 cpu_xnnpack: pass · 32.452 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
yolact-resnet50 | image-segmentation | — | pass · 1700.407 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | pass · 63.108 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | galaxy-s26 gpu_mldrift: pass · 38.367 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 · 2026-08-26galaxy-s26 npu_qnn: pass · 14.344 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · QAIRT (Hexagon, JIT on-device) · Android 16 · 2026-08-26raspberry-pi-5 cpu_xnnpack: pass · 1095.56 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
yunet-face | object-detection | — | pass · 140.42 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | pass · 6.527 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-11 | galaxy-s26 gpu_mldrift: pass · 3.315 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · Android 16 · 2026-08-25galaxy-s26 npu_qnn: pass · 1.85 ms stale — latest known 2.3.0Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert 2.2.0 · QAIRT (Hexagon, JIT on-device) · Android 16 · 2026-08-25raspberry-pi-5 cpu_xnnpack: pass · 33.696 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
zamba2-1.2b-instruct-int8 | text-generation | — | not measured | not measured | galaxy-s26 cpu: fallback · 12.21 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-09-05galaxy-s26 gpu: pass · 11.68 tok/s decode stale — latest known 0.19.1Galaxy S26 (SM-S942Q) · Qualcomm SM8850 · litert-lm 0.16.0 · Android 16 · 2026-08-24raspberry-pi-5 cpu: pass · 1.81 tok/s decode stale — latest known 0.19.1Raspberry Pi 5 Model B Rev 1.1 · Broadcom BCM2712 · litert-lm 0.16.1 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-09-02 | CARD.md · card.json | — |
zamba2-2.7b-instruct | text-generation | — | not measured | not measured | raspberry-pi-5 cpu: pass · 1.03 tok/s decode stale — latest known 0.19.1Raspberry Pi 5 Model B Rev 1.1 · Broadcom BCM2712 · litert-lm 0.16.1 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-09-02 | CARD.md · card.json | — |
zipformer-medium-cr-ctc__zipformer_ctc_fp16 | automatic-speech-recognition | — | pass · 2190.745 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-20 | output_mismatch · 25.19 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-20 | raspberry-pi-5 cpu_xnnpack: pass · 762.79 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
zipformer-medium-cr-ctc__zipformer_ctc_large_fp16 | automatic-speech-recognition | — | pass · 3691.958 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-20 | output_mismatch · 41.163 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-20 | raspberry-pi-5 cpu_xnnpack: pass · 1055.052 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
zipformer-medium-cr-ctc__zipformer_ctc_small_fp16 | automatic-speech-recognition | — | pass · 1433.62 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-20 | output_mismatch · 14.132 ms stale — latest known 2.6.0mac-studio-m4-max · chromium 151.0.7922.34 headless · macOS · @litertjs/core 2.5.3 · 2026-08-20 | raspberry-pi-5 cpu_xnnpack: pass · 606.175 msRaspberry Pi 5 Model B Rev 1.1 · litert 2.2.0.dev20260804 · Linux-6.18.34+rpt-rpi-2712-aarch64-with-glibc2.41 · 2026-08-31 | CARD.md · card.json | run it |
One-line prompt for Claude Code / Gemini CLI — re-runs the sweep behind this table on the committed example catalog, locally:
In the edge-compat repo, re-run the LiteRT.js sweep behind this table on the committed example catalog: cd web/sweep && npm ci && npm run sweep -- --catalog ../../data/examples/web_catalog_example.csv --out out/ — then compare the records under out/ with the example rows in this table.