{
  "artifacts": [
    {
      "file": "00519d5f6f83208e8a35817fc827a234e8070ed07dcb967aca06c40129bf4c21.tflite",
      "sha256": "6025e85bf481b9fb370dcd76b6ff5154d8958a80ab0e91521c3d1babf0c218f7",
      "size_mb": 840.879
    }
  ],
  "benchmarks": [],
  "browser": {
    "backends": [
      {
        "backend": "wasm_xnnpack",
        "date": "2026-08-11",
        "env": {
          "browser": "chromium",
          "browser_version": "151.0.7922.34",
          "headless": true,
          "jspi": true,
          "litertjs_core_version": "2.5.3",
          "machine_label": "mac-studio-m4-max",
          "os": "macOS",
          "os_version": "27.0.0",
          "webgpu_adapter": {
            "architecture": "metal-3",
            "description": "",
            "device": "",
            "vendor": "apple"
          }
        },
        "full_delegation": null,
        "latency_p50_ms": null,
        "loads": false,
        "max_rel_diff": null,
        "output_match": null,
        "provenance": "measured",
        "runs": false
      },
      {
        "backend": "webgpu_mldrift",
        "date": "2026-08-11",
        "env": {
          "browser": "chromium",
          "browser_version": "151.0.7922.34",
          "headless": true,
          "jspi": true,
          "litertjs_core_version": "2.5.3",
          "machine_label": "mac-studio-m4-max",
          "os": "macOS",
          "os_version": "27.0.0",
          "webgpu_adapter": {
            "architecture": "metal-3",
            "description": "",
            "device": "",
            "vendor": "apple"
          }
        },
        "full_delegation": null,
        "latency_p50_ms": null,
        "loads": false,
        "max_rel_diff": null,
        "output_match": null,
        "provenance": "measured",
        "runs": false
      }
    ],
    "demo_url": null,
    "sweep_source": "data/sweep/2.5.3/2026-08-11/qwen3-embedding-0.6b.json"
  },
  "conversion": {
    "command": "python text-embedding/conversion/build_qwen3emb.py (embedding table via export_embeddings.py)",
    "quantization": "fp16",
    "tool": "litert-torch",
    "tool_version": "0.10.0 (editable dev checkout 115a136 + local patches)"
  },
  "cross_runtime": [],
  "delegation": null,
  "device": {
    "records": [
      {
        "device": "galaxy-s26",
        "run": {
          "accelerator": "gpu_mldrift",
          "context_length": null,
          "date": "2026-08-27",
          "decode_tokens_per_s": null,
          "delegated_ops": null,
          "env": {
            "device": "Galaxy S26 (SM-S942Q)",
            "machine_label": "galaxy-s26-npubench",
            "os_build": "Android 16",
            "runtime": "litert",
            "runtime_version": "2.2.0",
            "soc": "Qualcomm SM8850",
            "vendor_sdk": null
          },
          "error": null,
          "evidence": [
            "npubench sweep row: N=50 median, 1 backend = 1 process, accepted only with thermal NONE->NONE; NPU rows additionally required qnn_partition delegate evidence in logcat (a stock model asked for on the NPU can silently land on XNNPACK and report a CPU number)",
            "mode=jit thermal=NONE->NONE headroom=0.7699999->0.77666664 attempt=0"
          ],
          "failure_class": null,
          "full_delegation": null,
          "latency_p50_ms": 44.738,
          "loads": true,
          "max_abs_diff": null,
          "max_rel_diff": null,
          "metrics": {
            "iterations": 50,
            "latency_max_ms": 49.237,
            "latency_min_ms": 43.751,
            "load_ms": 3842.3
          },
          "output_match": null,
          "peak_mem_mb": null,
          "prefill_tokens_per_s": null,
          "provenance": "measured",
          "runs": true,
          "total_ops": null,
          "ttft_ms": null
        },
        "source": "data/device_runs/2.2.0/2026-08-27/qwen3-embedding-0.6b__galaxy-s26.json"
      },
      {
        "device": "galaxy-s26",
        "run": {
          "accelerator": "npu_qnn",
          "context_length": null,
          "date": "2026-08-27",
          "decode_tokens_per_s": null,
          "delegated_ops": null,
          "env": {
            "device": "Galaxy S26 (SM-S942Q)",
            "machine_label": "galaxy-s26-npubench",
            "os_build": "Android 16",
            "runtime": "litert",
            "runtime_version": "2.2.0",
            "soc": "Qualcomm SM8850",
            "vendor_sdk": "QAIRT (Hexagon, AOT host-compiled, SM8850 target)"
          },
          "error": null,
          "evidence": [
            "npubench sweep row: N=50 median, 1 backend = 1 process, accepted only with thermal NONE->NONE; NPU rows additionally required qnn_partition delegate evidence in logcat (a stock model asked for on the NPU can silently land on XNNPACK and report a CPU number)",
            "mode=aot thermal=NONE->NONE headroom=0.7633334->0.7633334 attempt=0"
          ],
          "failure_class": null,
          "full_delegation": null,
          "latency_p50_ms": 25.44,
          "loads": true,
          "max_abs_diff": null,
          "max_rel_diff": null,
          "metrics": {
            "iterations": 50,
            "latency_max_ms": 25.975,
            "latency_min_ms": 25.059,
            "load_ms": 292.0,
            "qnn_partition_lines": 57
          },
          "output_match": null,
          "peak_mem_mb": null,
          "prefill_tokens_per_s": null,
          "provenance": "measured",
          "runs": true,
          "total_ops": null,
          "ttft_ms": null
        },
        "source": "data/device_runs/2.2.0/2026-08-27/qwen3-embedding-0.6b__galaxy-s26.json"
      }
    ]
  },
  "model": {
    "family": "qwen3",
    "id": "qwen3-embedding-0.6b",
    "license": "apache-2.0",
    "source_url": "https://huggingface.co/litert-community/Qwen3-Embedding-0.6B-LiteRT",
    "task": "sentence-similarity"
  },
  "pitfalls": [
    "Token embedding lookup is host-side from the shipped embeddings_fp16.bin table (GATHER is GPU-banned); the graph consumes inputs_embeds [1,128,1024] (zoo README L1830-1831; HF card)."
  ],
  "schema_version": "1.2"
}