{
  "artifacts": [
    {
      "file": "LFM2-2.6B-Exp_int8.litertlm",
      "sha256": "a44c3cfd0644e9fe9036c0efe3955df48a91990fe6e42855a9de0020c4e76258",
      "size_mb": 2597.627
    }
  ],
  "benchmarks": [],
  "conversion": {
    "command": "LiquidAI/LFM2-2.6B-Exp rev 2251c865: unquantized export (11 prefill signatures, 1024 down to 1, plus decode, over a 4096-position KV cache) -> post-hoc int8 quantization -> metadata repack with LiteRT-LM's lfm2 configuration (card L111-L112)",
    "quantization": "int8 dynamic on linears + int8 embedding, convs float (card L27)",
    "tool": "litert-torch 0.9.4; repro = hf-to-litertlm lfm_work/convert_lfm2_26b_exp.sh (card L111)",
    "tool_version": "litert-torch 0.9.4, litert-converter 0.4.0, ai-edge-quantizer 0.9.0, transformers 5.14.1, torch 2.13.0; packed by the litert-lm 0.18.0 CLI"
  },
  "cross_runtime": [],
  "delegation": null,
  "device": {
    "records": [
      {
        "device": "galaxy-s26",
        "run": {
          "accelerator": "cpu",
          "context_length": null,
          "date": "2026-10-12",
          "decode_tokens_per_s": 18.545,
          "delegated_ops": 1394,
          "env": {
            "device": "Galaxy S26 (SM-S942Q)",
            "machine_label": "galaxy-s26-aab1",
            "os_build": "Android 16 (BP4A.251205.006)",
            "runtime": "litert-lm",
            "runtime_version": "prebuilt-adac974b",
            "soc": "Qualcomm SM8850",
            "vendor_sdk": null
          },
          "error": null,
          "evidence": [
            "binary: Google's prebuilt litert_lm_advanced_main + 6 .so (gs://litert/binaries/latest/android_arm64/litert_lm/, object dated 2026-09-18, 40,382,152 B, sha256 adac974b…)",
            "r1 hold 2026-10-12 21:47:56-22:07:53 JST: [21:49:18] bundle on device: a44c3cfd0644e9fe9036c0efe3955df48a91990fe6e42855a9de0020c4e76258 (2723809744 B, push 72.6 s) = Mac",
            "bench args: --backend=cpu --model_path=/data/local/tmp/aa4_26bexp/LFM2-2.6B-Exp_int8.litertlm --benchmark=true --benchmark_prefill_tokens=1024 --benchmark_decode_tokens=256 --max_num_tokens=2048 --num_iterations=3 --report_peak_memory_footprint=true --metric_proto_file_path=/data/local/tmp/aa4_26bexp/metrics.pb",
            "VERBOSE: Replacing 869 out of 978 node(s) with delegate (TfLiteXNNPackDelegate) node, yielding 152 partitions for subgraph 0 (prefill_1024).",
            "VERBOSE: Replacing 795 out of 895 node(s) with delegate (TfLiteXNNPackDelegate) node, yielding 156 partitions for subgraph 11 (decode).",
            "VERBOSE: Replacing 1394 out of 1428 node(s) with delegate (TfLiteXNNPackDelegate) node, yielding 18 partitions for subgraph 0 (prefill_1024).",
            "VERBOSE: Replacing 1273 out of 1296 node(s) with delegate (TfLiteXNNPackDelegate) node, yielding 18 partitions for subgraph 11 (decode).",
            "[21:54:03] parsed: s26_i8_cpu_timing cpu: gate PASS 1/1 | bench prefill 285.63 decode 18.545 init 224.89 cold 3121.6 peak 4163.8 capped 23/24 | answers: 'Paris.'",
            "timing run prompt process: \"Paris.\"",
            "check run (hold 2026-10-12 21:47:56-22:07:53 JST): 2 of 2 prompt processes answered with the expected word (\"Paris.\")"
          ],
          "failure_class": null,
          "full_delegation": false,
          "latency_p50_ms": null,
          "loads": true,
          "max_abs_diff": null,
          "max_rel_diff": null,
          "metrics": {
            "boot_epoch_s": 1791590919.25,
            "capped_samples": 23.0,
            "cold_init_ms": 3121.6,
            "cpu_capped_samples": 23.0,
            "decode_tok_s_it1": 20.24,
            "decode_tok_s_it2": 17.75,
            "decode_tok_s_it3": 19.34,
            "decode_tokens": 256.0,
            "init_ms": 224.89,
            "iterations": 3.0,
            "kgsl_ceiling_min_mhz": 1300.0,
            "kgsl_temp_max_c": 65.9,
            "max_num_tokens": 2048.0,
            "memavail_at_bench_start_kib": 7082256.0,
            "memavail_min_kib": 5463848.0,
            "prefill_tok_s_it1": 405.6,
            "prefill_tok_s_it2": 295.01,
            "prefill_tok_s_it3": 276.25,
            "prefill_tokens": 1024.0,
            "sampler_samples": 24.0,
            "skin_c_at_bench_start": 37.9,
            "t_1024_256_s": 17.3893,
            "thermal_status_at_bench_start": 0.0,
            "ttft_s_it1": 2.57,
            "ttft_s_it2": 3.53,
            "ttft_s_it3": 3.76
          },
          "output_match": null,
          "peak_mem_mb": 4163.8,
          "prefill_tokens_per_s": 285.63,
          "provenance": "measured",
          "runs": true,
          "total_ops": 1428,
          "ttft_ms": 3645.0
        },
        "source": "data/device_runs/prebuilt-adac974b/2026-10-12/lfm2-2.6b-exp-int8__galaxy-s26.json"
      },
      {
        "device": "galaxy-s26",
        "run": {
          "accelerator": "gpu",
          "context_length": null,
          "date": "2026-10-12",
          "decode_tokens_per_s": 20.435,
          "delegated_ops": 978,
          "env": {
            "device": "Galaxy S26 (SM-S942Q)",
            "machine_label": "galaxy-s26-aab1",
            "os_build": "Android 16 (BP4A.251205.006)",
            "runtime": "litert-lm",
            "runtime_version": "prebuilt-adac974b",
            "soc": "Qualcomm SM8850",
            "vendor_sdk": null
          },
          "error": null,
          "evidence": [
            "binary: Google's prebuilt litert_lm_advanced_main + 6 .so (gs://litert/binaries/latest/android_arm64/litert_lm/, object dated 2026-09-18, 40,382,152 B, sha256 adac974b…)",
            "r1 hold 2026-10-12 21:47:56-22:07:53 JST: [22:00:24] bundle on device: a44c3cfd0644e9fe9036c0efe3955df48a91990fe6e42855a9de0020c4e76258 (2723809744 B, push 0 s) = Mac",
            "bench args: --backend=gpu --model_path=/data/local/tmp/aa4_26bexp/LFM2-2.6B-Exp_int8.litertlm --benchmark=true --benchmark_prefill_tokens=1024 --benchmark_decode_tokens=256 --max_num_tokens=2048 --num_iterations=3 --report_peak_memory_footprint=true --metric_proto_file_path=/data/local/tmp/aa4_26bexp/metrics.pb",
            "VERBOSE: Replacing 978 out of 978 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 0 (prefill_1024).",
            "VERBOSE: Replacing 978 out of 978 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 1 (prefill_512).",
            "VERBOSE: Replacing 978 out of 978 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 2 (prefill_256).",
            "VERBOSE: Replacing 978 out of 978 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 3 (prefill_128).",
            "VERBOSE: Replacing 978 out of 978 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 4 (prefill_64).",
            "VERBOSE: Replacing 978 out of 978 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 5 (prefill_32).",
            "VERBOSE: Replacing 978 out of 978 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 6 (prefill_16).",
            "VERBOSE: Replacing 978 out of 978 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 7 (prefill_8).",
            "VERBOSE: Replacing 978 out of 978 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 8 (prefill_4).",
            "VERBOSE: Replacing 978 out of 978 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 9 (prefill_2).",
            "VERBOSE: Replacing 877 out of 877 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 10 (prefill_1).",
            "VERBOSE: Replacing 895 out of 895 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 11 (decode).",
            "[22:03:22] parsed: s26_i8_gpu_timing gpu: gate PASS 1/1 | bench prefill 1140.525 decode 20.435000000000002 init 4761.94 cold 8659.3 peak 779.434 capped 18/22 | answers: 'Paris.'",
            "timing run prompt process: \"Paris.\"",
            "check run (hold 2026-10-12 21:47:56-22:07:53 JST): 2 of 2 prompt processes answered with the expected word (\"Paris.\")"
          ],
          "failure_class": null,
          "full_delegation": true,
          "latency_p50_ms": null,
          "loads": true,
          "max_abs_diff": null,
          "max_rel_diff": null,
          "metrics": {
            "boot_epoch_s": 1791590918.43,
            "capped_samples": 18.0,
            "cold_init_ms": 8659.3,
            "cpu_capped_samples": 19.0,
            "decode_tok_s_it1": 20.1,
            "decode_tok_s_it2": 20.59,
            "decode_tok_s_it3": 20.28,
            "decode_tokens": 256.0,
            "init_ms": 4761.94,
            "iterations": 3.0,
            "kgsl_ceiling_min_mhz": 826.0,
            "kgsl_temp_max_c": 90.3,
            "max_num_tokens": 2048.0,
            "memavail_at_bench_start_kib": 7434972.0,
            "memavail_min_kib": 3427232.0,
            "prefill_tok_s_it1": 1262.7,
            "prefill_tok_s_it2": 1190.16,
            "prefill_tok_s_it3": 1090.89,
            "prefill_tokens": 1024.0,
            "sampler_samples": 22.0,
            "skin_c_at_bench_start": 42.0,
            "t_1024_256_s": 13.4254,
            "thermal_status_at_bench_start": 2.0,
            "ttft_s_it1": 0.86,
            "ttft_s_it2": 0.91,
            "ttft_s_it3": 0.99
          },
          "output_match": null,
          "peak_mem_mb": 779.434,
          "prefill_tokens_per_s": 1140.525,
          "provenance": "measured",
          "runs": true,
          "total_ops": 978,
          "ttft_ms": 950.0
        },
        "source": "data/device_runs/prebuilt-adac974b/2026-10-12/lfm2-2.6b-exp-int8__galaxy-s26.json"
      },
      {
        "device": "mac-studio-m4-max",
        "run": {
          "accelerator": "cpu",
          "context_length": null,
          "date": "2026-10-13",
          "decode_tokens_per_s": null,
          "delegated_ops": null,
          "env": {
            "device": "Mac Studio (M4 Max)",
            "machine_label": "mac-studio-m4-max",
            "os_build": "macOS 27.0",
            "runtime": "litert-lm",
            "runtime_version": "0.17.1",
            "soc": "Apple M4 Max",
            "vendor_sdk": null
          },
          "error": "It seems like your message got cut off. Could you please provide more details or clarify what you'd like assistance with? Whether it's a question, a topic you're curious about, or something else, I'm here to help!",
          "evidence": [
            "command: litert-lm run LFM2-2.6B-Exp_int8.litertlm --prompt \"What is the capital of France?\" --cache no < /dev/null",
            "litert-lm 0.17.1 (the log's version line)",
            "stdout: It seems like your message got cut off. Could you please provide more details or clarify what you'd like assistance with? Whether it's a question, a topic you're curious about, or something else, I'm here to help! Here are some examples of what you might be asking about: - **General Knowledge**: Facts, definitions, or explanations. - **Problem Solving**: Math problems, coding, or logical puzzles. - **Writing Help**: Editing, brainstorming, or structuring content. - **Learning Tips**: Study strategies, productivity, or skill development. - **Creative Ideas**: Writing prompts, project concepts, or brainstorming. Let me know how I can assist! 😊",
            "### EXIT: 0",
            "2026-10-13 10:02:54 JST 0.17.1 run int8: It seems like your message got cut off. Could you please provide more details or clarify what you'd like assistance with? Whether it's a question, a topic you're curious about, or something else, I'm here to help!  Here are some examples of what you might be "
          ],
          "failure_class": "degenerate_output",
          "full_delegation": null,
          "latency_p50_ms": null,
          "loads": true,
          "max_abs_diff": null,
          "max_rel_diff": null,
          "metrics": {
            "processes": 1.0
          },
          "output_match": null,
          "peak_mem_mb": null,
          "prefill_tokens_per_s": null,
          "provenance": "measured",
          "runs": false,
          "total_ops": null,
          "ttft_ms": null
        },
        "source": "data/device_runs/0.17.1/2026-10-13/lfm2-2.6b-exp-int8__mac-studio-m4-max.json"
      },
      {
        "device": "mac-studio-m4-max",
        "run": {
          "accelerator": "cpu",
          "context_length": null,
          "date": "2026-10-12",
          "decode_tokens_per_s": 38.27,
          "delegated_ops": null,
          "env": {
            "device": "Mac Studio (M4 Max)",
            "machine_label": "mac-studio-m4-max",
            "os_build": "macOS 27.0",
            "runtime": "litert-lm",
            "runtime_version": "0.18.0",
            "soc": "Apple M4 Max",
            "vendor_sdk": null
          },
          "error": null,
          "evidence": [
            "Prefill 377.26 / decode 38.27 tok/s, TTFT 2.7405 s, init 31.9277 s",
            "settle_run: quiet after 12 s; run: litert-lm benchmark LFM2-2.6B-Exp_int8.litertlm --backend cpu -p 1024 -d 256 --max-num-tokens 2048 --runs 3 --cache no",
            "2026-10-12 21:18:43 JST cpu int8 exit 0"
          ],
          "failure_class": null,
          "full_delegation": null,
          "latency_p50_ms": null,
          "loads": true,
          "max_abs_diff": null,
          "max_rel_diff": null,
          "metrics": {
            "decode_tokens": 256.0,
            "init_s": 31.9277,
            "iterations": 3.0,
            "max_num_tokens": 2048.0,
            "prefill_tokens": 1024.0,
            "processes": 1.0,
            "t_1024_256_s": 9.4036
          },
          "output_match": null,
          "peak_mem_mb": null,
          "prefill_tokens_per_s": 377.26,
          "provenance": "measured",
          "runs": true,
          "total_ops": null,
          "ttft_ms": 2740.5
        },
        "source": "data/device_runs/0.18.0/2026-10-12/lfm2-2.6b-exp-int8__mac-studio-m4-max.json"
      },
      {
        "device": "mac-studio-m4-max",
        "run": {
          "accelerator": "gpu",
          "context_length": null,
          "date": "2026-10-12",
          "decode_tokens_per_s": 123.39,
          "delegated_ops": null,
          "env": {
            "device": "Mac Studio (M4 Max)",
            "machine_label": "mac-studio-m4-max",
            "os_build": "macOS 27.0",
            "runtime": "litert-lm",
            "runtime_version": "0.18.0",
            "soc": "Apple M4 Max",
            "vendor_sdk": null
          },
          "error": null,
          "evidence": [
            "Prefill 2441.36 / decode 123.39 tok/s, TTFT 0.4275 s, init 2.8241 s",
            "settle_run: quiet after 3 s; run: litert-lm benchmark LFM2-2.6B-Exp_int8.litertlm --backend gpu -p 1024 -d 256 --max-num-tokens 2048 --runs 3 --cache no",
            "2026-10-12 21:30:04 JST gpu int8 verbose exit 0 validation_error=0 shape_mismatch=0 replacing=24",
            "2026-10-12 21:28:57 JST gpu int8 exit 0"
          ],
          "failure_class": null,
          "full_delegation": null,
          "latency_p50_ms": null,
          "loads": true,
          "max_abs_diff": null,
          "max_rel_diff": null,
          "metrics": {
            "decode_tokens": 256.0,
            "init_s": 2.8241,
            "iterations": 3.0,
            "max_num_tokens": 2048.0,
            "prefill_tokens": 1024.0,
            "processes": 1.0,
            "t_1024_256_s": 2.4942
          },
          "output_match": null,
          "peak_mem_mb": null,
          "prefill_tokens_per_s": 2441.36,
          "provenance": "measured",
          "runs": true,
          "total_ops": null,
          "ttft_ms": 427.5
        },
        "source": "data/device_runs/0.18.0/2026-10-12/lfm2-2.6b-exp-int8__mac-studio-m4-max.json"
      },
      {
        "device": "pixel-8a",
        "run": {
          "accelerator": "cpu",
          "context_length": null,
          "date": "2026-10-12",
          "decode_tokens_per_s": 5.845,
          "delegated_ops": 1394,
          "env": {
            "device": "Pixel 8a",
            "machine_label": "pixel-8a-aab1",
            "os_build": "Android 16 (CP1A.260505.005)",
            "runtime": "litert-lm",
            "runtime_version": "prebuilt-adac974b",
            "soc": "Tensor G3",
            "vendor_sdk": null
          },
          "error": null,
          "evidence": [
            "binary: Google's prebuilt litert_lm_advanced_main + 6 .so (gs://litert/binaries/latest/android_arm64/litert_lm/, object dated 2026-09-18, 40,382,152 B, sha256 adac974b…)",
            "r1 hold 2026-10-12 21:46:00-22:12:37 JST: [21:47:12] bundle on device: a44c3cfd0644e9fe9036c0efe3955df48a91990fe6e42855a9de0020c4e76258 (2723809744 B, push 58.9 s) = Mac",
            "bench args: --backend=cpu --model_path=/data/local/tmp/aa4_26bexp/LFM2-2.6B-Exp_int8.litertlm --benchmark=true --benchmark_prefill_tokens=1024 --benchmark_decode_tokens=256 --max_num_tokens=2048 --num_iterations=3 --report_peak_memory_footprint=true --metric_proto_file_path=/data/local/tmp/aa4_26bexp/metrics.pb",
            "VERBOSE: Replacing 869 out of 978 node(s) with delegate (TfLiteXNNPackDelegate) node, yielding 152 partitions for subgraph 0 (prefill_1024).",
            "VERBOSE: Replacing 795 out of 895 node(s) with delegate (TfLiteXNNPackDelegate) node, yielding 156 partitions for subgraph 11 (decode).",
            "VERBOSE: Replacing 1394 out of 1428 node(s) with delegate (TfLiteXNNPackDelegate) node, yielding 18 partitions for subgraph 0 (prefill_1024).",
            "VERBOSE: Replacing 1273 out of 1296 node(s) with delegate (TfLiteXNNPackDelegate) node, yielding 18 partitions for subgraph 11 (decode).",
            "[21:54:32] parsed: p8a_i8_cpu_timing cpu: gate PASS 1/1 | bench prefill 70.19 decode 5.845 init 708.62 cold 11002.71 peak 3659.46 capped 51/53 | answers: 'Paris.'",
            "timing run prompt process: \"Paris.\"",
            "check run (hold 2026-10-12 21:46:00-22:12:37 JST): 1 of 1 prompt processes answered with the expected word (\"Paris.\")"
          ],
          "failure_class": null,
          "full_delegation": false,
          "latency_p50_ms": null,
          "loads": true,
          "max_abs_diff": null,
          "max_rel_diff": null,
          "metrics": {
            "boot_epoch_s": 1791442100.04,
            "capped_samples": 51.0,
            "cold_init_ms": 11002.71,
            "cpu_capped_samples": 51.0,
            "decode_tok_s_it1": 8.53,
            "decode_tok_s_it2": 5.81,
            "decode_tok_s_it3": 5.88,
            "decode_tokens": 256.0,
            "init_ms": 708.62,
            "iterations": 3.0,
            "max_num_tokens": 2048.0,
            "memavail_at_bench_start_kib": 4071792.0,
            "memavail_min_kib": 3082320.0,
            "prefill_tok_s_it1": 106.8,
            "prefill_tok_s_it2": 79.9,
            "prefill_tok_s_it3": 60.48,
            "prefill_tokens": 1024.0,
            "sampler_samples": 53.0,
            "t_1024_256_s": 58.3871,
            "thermal_status_at_bench_start": 1.0,
            "ttft_s_it1": 9.71,
            "ttft_s_it2": 12.99,
            "ttft_s_it3": 17.1
          },
          "output_match": null,
          "peak_mem_mb": 3659.46,
          "prefill_tokens_per_s": 70.19,
          "provenance": "measured",
          "runs": true,
          "total_ops": 1428,
          "ttft_ms": 15045.0
        },
        "source": "data/device_runs/prebuilt-adac974b/2026-10-12/lfm2-2.6b-exp-int8__pixel-8a.json"
      },
      {
        "device": "pixel-8a",
        "run": {
          "accelerator": "gpu",
          "context_length": null,
          "date": "2026-10-12",
          "decode_tokens_per_s": 11.095,
          "delegated_ops": 978,
          "env": {
            "device": "Pixel 8a",
            "machine_label": "pixel-8a-aab1",
            "os_build": "Android 16 (CP1A.260505.005)",
            "runtime": "litert-lm",
            "runtime_version": "prebuilt-adac974b",
            "soc": "Tensor G3",
            "vendor_sdk": null
          },
          "error": null,
          "evidence": [
            "binary: Google's prebuilt litert_lm_advanced_main + 6 .so (gs://litert/binaries/latest/android_arm64/litert_lm/, object dated 2026-09-18, 40,382,152 B, sha256 adac974b…)",
            "r3 hold 2026-10-12 22:40:18-23:09:17 JST: [22:52:10] bundle on device: a44c3cfd0644e9fe9036c0efe3955df48a91990fe6e42855a9de0020c4e76258 (2723809744 B, push 0 s) = Mac",
            "bench args: --backend=gpu --model_path=/data/local/tmp/aa4_26bexp/LFM2-2.6B-Exp_int8.litertlm --benchmark=true --benchmark_prefill_tokens=1024 --benchmark_decode_tokens=256 --max_num_tokens=2048 --num_iterations=3 --report_peak_memory_footprint=true --metric_proto_file_path=/data/local/tmp/aa4_26bexp/metrics.pb",
            "VERBOSE: Replacing 978 out of 978 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 0 (prefill_1024).",
            "VERBOSE: Replacing 978 out of 978 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 1 (prefill_512).",
            "VERBOSE: Replacing 978 out of 978 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 2 (prefill_256).",
            "VERBOSE: Replacing 978 out of 978 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 3 (prefill_128).",
            "VERBOSE: Replacing 978 out of 978 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 4 (prefill_64).",
            "VERBOSE: Replacing 978 out of 978 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 5 (prefill_32).",
            "VERBOSE: Replacing 978 out of 978 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 6 (prefill_16).",
            "VERBOSE: Replacing 978 out of 978 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 7 (prefill_8).",
            "VERBOSE: Replacing 978 out of 978 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 8 (prefill_4).",
            "VERBOSE: Replacing 978 out of 978 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 9 (prefill_2).",
            "VERBOSE: Replacing 877 out of 877 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 10 (prefill_1).",
            "VERBOSE: Replacing 895 out of 895 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 11 (decode).",
            "[23:00:14] parsed: p8a_i8_gpu_timing gpu: gate PASS 1/1 | bench prefill 498.54499999999996 decode 11.094999999999999 init 12495.64 cold 106177.27 peak 1667.16 capped 0/34 | answers: 'Paris.'",
            "timing run prompt process: \"Paris.\"",
            "check run (hold 2026-10-12 22:40:18-23:09:17 JST): 1 of 1 prompt processes answered with the expected word (\"Paris.\")"
          ],
          "failure_class": null,
          "full_delegation": true,
          "latency_p50_ms": null,
          "loads": true,
          "max_abs_diff": null,
          "max_rel_diff": null,
          "metrics": {
            "boot_epoch_s": 1791442100.46,
            "capped_samples": 0.0,
            "cold_init_ms": 106177.27,
            "cpu_capped_samples": 26.0,
            "decode_tok_s_it1": 10.86,
            "decode_tok_s_it2": 11.28,
            "decode_tok_s_it3": 10.91,
            "decode_tokens": 256.0,
            "init_ms": 12495.64,
            "iterations": 3.0,
            "max_num_tokens": 2048.0,
            "memavail_at_bench_start_kib": 4338712.0,
            "memavail_min_kib": 501160.0,
            "prefill_tok_s_it1": 509.27,
            "prefill_tok_s_it2": 501.58,
            "prefill_tok_s_it3": 495.51,
            "prefill_tokens": 1024.0,
            "sampler_samples": 34.0,
            "t_1024_256_s": 25.1274,
            "thermal_status_at_bench_start": 1.0,
            "ttft_s_it1": 2.1,
            "ttft_s_it2": 2.13,
            "ttft_s_it3": 2.16
          },
          "output_match": null,
          "peak_mem_mb": 1667.16,
          "prefill_tokens_per_s": 498.545,
          "provenance": "measured",
          "runs": true,
          "total_ops": 978,
          "ttft_ms": 2145.0
        },
        "source": "data/device_runs/prebuilt-adac974b/2026-10-12/lfm2-2.6b-exp-int8__pixel-8a.json"
      }
    ]
  },
  "model": {
    "family": "lfm2",
    "id": "lfm2-2.6b-exp-int8",
    "license": "other (lfm1.0)",
    "source_url": "https://huggingface.co/litert-community/LFM2-2.6B-Exp",
    "task": "text-generation"
  },
  "pitfalls": [
    "Runtime: tested on litert-lm 0.18.0 on a Mac; on litert-lm 0.17.1 this file answered \"What is the capital of France?\" as if no question had been asked (one CPU run) (card L21).",
    "Reasoning trace: Liquid AI's card says nothing about reasoning, yet the model sometimes writes a trace between <think> and </think>, which goes to the thought channel (printed by the CLI between [thought] and [/thought]); one appeared for \"What is the capital of France?\" (this file, CPU and GPU; with a 16-token output limit the reply was empty, all 16 tokens being trace) and for 1 of 120 IFEval-lite prompts; --thinking-budget 0 and 32 did not suppress it (card L23, L72).",
    "Template: Liquid AI's chat template (the LFM2-2.6B file) fails at render on litert-lm 0.18.0 (map has no method named get), so the bundle carries LiteRT-LM's lfm2 template (byte-identical to models/lfm2/chat_template.jinja at commit 86a7e6d0), which reads message content as a list of parts; stops <|im_end|> and <|endoftext|>, start token <|startoftext|> (card L72, L113).",
    "Tool calling is declared (model type lfm2): the file returned a parsed get_weather call (Python API, CPU, greedy) and used its result with automatic tool calling; with tools the template omits Liquid AI's tool-list and tool-response markers, wraps the tool result in [ … ] and quotes call arguments with double quotes (card L74).",
    "One Engine per conversation: with the same system prompt in two conversations and user messages beginning with the same eight words (the condition of LiteRT-LM issue #3165), a shared Engine answered \"The third item in the list is cherry.\" where a fresh Engine answered \"cherry\" (3 of 3 each, both correct); with different system prompts the answers matched a fresh Engine (card L76).",
    "Conversation preface with tools and no messages: LiteRT-LM renders the chat template with an empty message list, and the template reads messages[0] without checking that the list is non-empty, so the render fails with 'Failed to apply template: undefined value' when the preface is prefilled at creation (prefillPrefaceOnInit, default false; creating the conversation fails) and in the preface render (renderPrefaceIntoString and its C++ and C API forms); the same with extra context (such as enable_thinking) and no messages, and in the preface render with an empty preface (run in the runtime on one Qwen3 template; the other templates rendered with minijinja 2.14 and jinja2); read in the source of LiteRT-LM 0.17.0, 0.18.1 and main, measured with the 0.18.1 PyPI package through its C API on these files' template text; the default path, where the first message is rendered with the preface, is not affected (the 0.18.1 Python API and CLI expose neither the flag nor the preface render); a system message or any initial message in the conversation config avoids it (card L117).",
    "Sampler: the bundle's default sampler is temperature 0.1, top-k 50, top-p 1.0; Liquid AI recommends temperature 0.3, min_p 0.15 and repetition_penalty 1.05, which the bundle does not carry; --top-k 1 gives greedy output (card L70).",
    "Cache: without --cache, `run` writes a compiled-weights cache next to the model (2.58 GB for this file); without --backend, `run` uses the CPU; on the Mac CPU this file ended normally on the dates prompt in 3 of 3 runs (card L68).",
    "Speed: int4 is the faster file on every GPU and this file on every CPU (Mac 9.40 vs 11.11 s, Galaxy S26 17.39 vs 25.91 s, Pixel 8a 58.42 vs 65.57 s for a 1024-token prefill plus 256 decoded tokens); not run on the NPU (card L32-L39, L107).",
    "iOS not measured: this file's one weights section is 2,722,843,088 bytes (2.54 GiB); the LFM2.5-2.6B card reports an int8 file with a 2.87 GB weights section running on an iPhone 17 Pro only with the extended-virtual-addressing and increased-memory entitlements (card L39).",
    "Phone conditions: on the Galaxy S26 the GPU clock ceiling fell from 1300 to 826 MHz inside this file's GPU benchmark and the CPU clock was lowered in 23 of 24 samples of its CPU run; on the Pixel 8a the CPU clock was lowered in 51 of 53 samples of the CPU run, and in the GPU benchmark the lowest MemAvailable sample was 501,160 kB (card L98, L107).",
    "Accuracy (greedy, at most 2048 output tokens, litert-lm on the Mac GPU, comparable only within the card's table; Liquid AI publishes no benchmark table for this checkpoint): GSM8K 87 of 100 (bf16 89), IFEval-lite 102 of 120 (bf16 102); 8-question check 7/8 on CPU and GPU, bf16 7/8, the same rhyme question missed (card L43, L49-L57).",
    "License: the LFM Open License v1.0's section 5 conditions commercial use on an annual revenue below 10 million US dollars (card L121)."
  ],
  "schema_version": "1.2"
}
