{
  "artifacts": [
    {
      "file": "LFM2-700M_int8.litertlm",
      "sha256": "ccc95007d868878ac749073cbc28b88097cb7d437fca3f9aec72f535cc7fd99b",
      "size_mb": 770.837
    }
  ],
  "benchmarks": [],
  "conversion": {
    "command": "LiquidAI/LFM2-700M rev 86f49fc9: export with litert-torch's default int8 recipe dynamic_wi8_afp32 (11 prefill signatures, 1024 down to 1, plus decode, over a 4096-position KV cache) -> metadata repack with LiteRT-LM's lfm2 configuration (card L110-L111)",
    "quantization": "int8 weights on linears, the 10 convolutions and the embedding (litert-torch's default export recipe dynamic_wi8_afp32), float activations (card L27)",
    "tool": "litert-torch 0.9.4; repro = hf-to-litertlm lfm_work/convert_lfm2_700m.sh (card L110)",
    "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": 45.65,
          "delegated_ops": 779,
          "env": {
            "device": "Galaxy S26 (SM-S942Q)",
            "machine_label": "galaxy-s26-aab1",
            "os_build": "Android 16",
            "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…)",
            "r3 hold 2026-10-12 21:40:50-21:47:53 JST: [21:44:42] bundle on device: ccc95007d868878ac749073cbc28b88097cb7d437fca3f9aec72f535cc7fd99b (808281472 B, push 0 s) = Mac",
            "bench args: --backend=cpu --model_path=/data/local/tmp/aa2_700m/LFM2-700M_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/aa2_700m/metrics.pb",
            "VERBOSE: Replacing 478 out of 560 node(s) with delegate (TfLiteXNNPackDelegate) node, yielding 106 partitions for subgraph 0 (prefill_1024).",
            "VERBOSE: Replacing 445 out of 519 node(s) with delegate (TfLiteXNNPackDelegate) node, yielding 112 partitions for subgraph 11 (decode).",
            "VERBOSE: Replacing 779 out of 818 node(s) with delegate (TfLiteXNNPackDelegate) node, yielding 32 partitions for subgraph 0 (prefill_1024).",
            "VERBOSE: Replacing 727 out of 756 node(s) with delegate (TfLiteXNNPackDelegate) node, yielding 34 partitions for subgraph 11 (decode).",
            "[21:47:52] parsed: i8_cpu_timing_rested cpu: gate PASS 1/1 | bench prefill 890.135 decode 45.650000000000006 init 125.15 cold 925.54 peak 1947.98 capped 9/10 | answers: 'The capital of France is Paris.'",
            "timing run prompt process: \"The capital of France is 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": 1791590918.94,
            "capped_samples": 9.0,
            "cold_init_ms": 925.54,
            "cpu_capped_samples": 9.0,
            "decode_tok_s_it1": 58.17,
            "decode_tok_s_it2": 46.56,
            "decode_tok_s_it3": 44.74,
            "decode_tokens": 256.0,
            "init_ms": 125.15,
            "iterations": 3.0,
            "kgsl_ceiling_min_mhz": 1300.0,
            "kgsl_temp_max_c": 65.1,
            "max_num_tokens": 2048.0,
            "memavail_at_bench_start_kib": 7047704.0,
            "memavail_min_kib": 5842388.0,
            "prefill_tok_s_it1": 977.68,
            "prefill_tok_s_it2": 1019.34,
            "prefill_tok_s_it3": 760.93,
            "prefill_tokens": 1024.0,
            "sampler_samples": 10.0,
            "skin_c_at_bench_start": 37.9,
            "t_1024_256_s": 6.7583,
            "thermal_status_at_bench_start": 0.0,
            "ttft_s_it1": 1.06,
            "ttft_s_it2": 1.03,
            "ttft_s_it3": 1.37
          },
          "output_match": null,
          "peak_mem_mb": 1947.98,
          "prefill_tokens_per_s": 890.135,
          "provenance": "measured",
          "runs": true,
          "total_ops": 818,
          "ttft_ms": 1200.0
        },
        "source": "data/device_runs/prebuilt-adac974b/2026-10-12/lfm2-700m-int8__galaxy-s26.json"
      },
      {
        "device": "galaxy-s26",
        "run": {
          "accelerator": "gpu",
          "context_length": null,
          "date": "2026-10-12",
          "decode_tokens_per_s": 58.965,
          "delegated_ops": 560,
          "env": {
            "device": "Galaxy S26 (SM-S942Q)",
            "machine_label": "galaxy-s26-aab1",
            "os_build": "Android 16",
            "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…)",
            "r3 hold 2026-10-12 21:40:50-21:47:53 JST: [21:41:15] bundle on device: ccc95007d868878ac749073cbc28b88097cb7d437fca3f9aec72f535cc7fd99b (808281472 B, push 21.2 s) = Mac",
            "bench args: --backend=gpu --model_path=/data/local/tmp/aa2_700m/LFM2-700M_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/aa2_700m/metrics.pb",
            "VERBOSE: Replacing 560 out of 560 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 0 (prefill_1024).",
            "VERBOSE: Replacing 560 out of 560 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 1 (prefill_512).",
            "VERBOSE: Replacing 560 out of 560 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 2 (prefill_256).",
            "VERBOSE: Replacing 560 out of 560 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 3 (prefill_128).",
            "VERBOSE: Replacing 560 out of 560 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 4 (prefill_64).",
            "VERBOSE: Replacing 560 out of 560 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 5 (prefill_32).",
            "VERBOSE: Replacing 560 out of 560 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 6 (prefill_16).",
            "VERBOSE: Replacing 560 out of 560 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 7 (prefill_8).",
            "VERBOSE: Replacing 560 out of 560 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 8 (prefill_4).",
            "VERBOSE: Replacing 560 out of 560 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 9 (prefill_2).",
            "VERBOSE: Replacing 501 out of 501 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 10 (prefill_1).",
            "VERBOSE: Replacing 519 out of 519 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 11 (decode).",
            "[21:42:45] parsed: i8_gpu_timing_rested gpu: gate PASS 1/1 | bench prefill 3396.395 decode 58.965 init 1276.15 cold 4304.28 peak 517.246 capped 0/8 | answers: 'The capital of France is Paris.'",
            "timing run prompt process: \"The capital of France is 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": 1791590919.25,
            "capped_samples": 0.0,
            "cold_init_ms": 4304.28,
            "cpu_capped_samples": 5.0,
            "decode_tok_s_it1": 55.79,
            "decode_tok_s_it2": 58.7,
            "decode_tok_s_it3": 59.23,
            "decode_tokens": 256.0,
            "init_ms": 1276.15,
            "iterations": 3.0,
            "kgsl_ceiling_min_mhz": 1300.0,
            "kgsl_temp_max_c": 76.0,
            "max_num_tokens": 2048.0,
            "memavail_at_bench_start_kib": 7071160.0,
            "memavail_min_kib": 5181592.0,
            "prefill_tok_s_it1": 3317.53,
            "prefill_tok_s_it2": 3479.72,
            "prefill_tok_s_it3": 3313.07,
            "prefill_tokens": 1024.0,
            "sampler_samples": 8.0,
            "skin_c_at_bench_start": 37.9,
            "t_1024_256_s": 4.6431,
            "thermal_status_at_bench_start": 0.0,
            "ttft_s_it1": 0.33,
            "ttft_s_it2": 0.31,
            "ttft_s_it3": 0.33
          },
          "output_match": null,
          "peak_mem_mb": 517.246,
          "prefill_tokens_per_s": 3396.395,
          "provenance": "measured",
          "runs": true,
          "total_ops": 560,
          "ttft_ms": 320.0
        },
        "source": "data/device_runs/prebuilt-adac974b/2026-10-12/lfm2-700m-int8__galaxy-s26.json"
      },
      {
        "device": "mac-studio-m4-max",
        "run": {
          "accelerator": "cpu",
          "context_length": null,
          "date": "2026-10-12",
          "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": "Hello! It seems like your message might have been cut off. Could you please provide more details so I can assist you better? Whether you're asking about a specific topic, need help with a question, or just want to chat, I'm here to help!",
          "evidence": [
            "command (tools/card_probes.sh; $I8 = LFM2-700M_int8.litertlm): litert-lm run \"$I8\" --prompt \"What is the capital of France?\" --cache no < /dev/null > \"$o\" 2>&1",
            "litert-lm 0.17.1",
            "stdout: Hello! It seems like your message might have been cut off. Could you please provide more details so I can assist you better? Whether you're asking about a specific topic, need help with a question, or just want to chat, I'm here to help!",
            "EXIT=0",
            "2026-10-12 20:14:21 JST 0.17.1 run exit 0: Hello! It seems like your message might have been cut off. Could you please provide more details so I can assist you better? Whether you're asking about a specific topic, need help with a question, or just want to chat, I'm here to help! EXIT=0"
          ],
          "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/0.17.1/2026-10-12/lfm2-700m-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": 90.87,
          "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 1129.14 / decode 88.59 tok/s, TTFT 0.9183 s, init 10.1786 s",
            "Prefill 1139.36 / decode 93.15 tok/s, TTFT 0.9095 s, init 9.0177 s",
            "2026-10-12 (cpu window) cpu wi8 p1 exit 0 Prefill speed: 1129.14 tokens/s Decode speed: 88.59 tokens/s",
            "2026-10-12 (cpu window) cpu wi8 p2 exit 0 Prefill speed: 1139.36 tokens/s Decode speed: 93.15 tokens/s"
          ],
          "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": 9.5982,
            "iterations": 3.0,
            "max_num_tokens": 2048.0,
            "prefill_tokens": 1024.0,
            "processes": 2.0,
            "run1_decode_tok_s": 88.59,
            "run1_init_s": 10.1786,
            "run1_prefill_tok_s": 1129.14,
            "run1_ttft_s": 0.9183,
            "run2_decode_tok_s": 93.15,
            "run2_init_s": 9.0177,
            "run2_prefill_tok_s": 1139.36,
            "run2_ttft_s": 0.9095,
            "t_1024_256_s": 3.72
          },
          "output_match": null,
          "peak_mem_mb": null,
          "prefill_tokens_per_s": 1134.25,
          "provenance": "measured",
          "runs": true,
          "total_ops": null,
          "ttft_ms": 913.9
        },
        "source": "data/device_runs/0.18.0/2026-10-12/lfm2-700m-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": 306.34,
          "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 7869.64 / decode 308.89 tok/s, TTFT 0.1334 s, init 1.2586 s",
            "Prefill 7841.43 / decode 305.24 tok/s, TTFT 0.1339 s, init 1.3163 s",
            "Prefill 7870.47 / decode 306.34 tok/s, TTFT 0.1334 s, init 1.2774 s",
            "2026-10-12 19:53:56 JST gpu wi8 verbose exit 0 validation_error=0 shape_mismatch=0 replacing=24",
            "2026-10-12 19:53:20 JST gpu wi8 exit 0",
            "2026-10-12 20:12:11 JST gpu wi8 p1 exit 0 Prefill speed: 7841.43 tokens/s Decode speed: 305.24 tokens/s",
            "2026-10-12 20:12:17 JST gpu wi8 p2 exit 0 Prefill speed: 7870.47 tokens/s Decode speed: 306.34 tokens/s"
          ],
          "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": 1.2774,
            "iterations": 3.0,
            "max_num_tokens": 2048.0,
            "prefill_tokens": 1024.0,
            "processes": 3.0,
            "run1_decode_tok_s": 308.89,
            "run1_init_s": 1.2586,
            "run1_prefill_tok_s": 7869.64,
            "run1_ttft_s": 0.1334,
            "run2_decode_tok_s": 305.24,
            "run2_init_s": 1.3163,
            "run2_prefill_tok_s": 7841.43,
            "run2_ttft_s": 0.1339,
            "run3_decode_tok_s": 306.34,
            "run3_init_s": 1.2774,
            "run3_prefill_tok_s": 7870.47,
            "run3_ttft_s": 0.1334,
            "t_1024_256_s": 0.9658
          },
          "output_match": null,
          "peak_mem_mb": null,
          "prefill_tokens_per_s": 7870.47,
          "provenance": "measured",
          "runs": true,
          "total_ops": null,
          "ttft_ms": 133.4
        },
        "source": "data/device_runs/0.18.0/2026-10-12/lfm2-700m-int8__mac-studio-m4-max.json"
      },
      {
        "device": "pixel-8a",
        "run": {
          "accelerator": "cpu",
          "context_length": null,
          "date": "2026-10-12",
          "decode_tokens_per_s": 22.725,
          "delegated_ops": 779,
          "env": {
            "device": "Pixel 8a",
            "machine_label": "pixel-8a-aab1",
            "os_build": "Android 16",
            "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 20:40:55-21:06:15 JST: [20:41:18] bundle on device: ccc95007d868878ac749073cbc28b88097cb7d437fca3f9aec72f535cc7fd99b (808281472 B, push 17.2 s) = Mac",
            "bench args: --backend=cpu --model_path=/data/local/tmp/aa2_700m/LFM2-700M_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/aa2_700m/metrics.pb",
            "VERBOSE: Replacing 478 out of 560 node(s) with delegate (TfLiteXNNPackDelegate) node, yielding 106 partitions for subgraph 0 (prefill_1024).",
            "VERBOSE: Replacing 445 out of 519 node(s) with delegate (TfLiteXNNPackDelegate) node, yielding 112 partitions for subgraph 11 (decode).",
            "VERBOSE: Replacing 779 out of 818 node(s) with delegate (TfLiteXNNPackDelegate) node, yielding 32 partitions for subgraph 0 (prefill_1024).",
            "VERBOSE: Replacing 727 out of 756 node(s) with delegate (TfLiteXNNPackDelegate) node, yielding 34 partitions for subgraph 11 (decode).",
            "[20:45:48] parsed: i8_cpu_timing cpu: gate PASS 1/1 | bench prefill 207.01 decode 22.725 init 455.25 cold 2026.65 peak 1534.82 capped 13/17 | answers: 'The capital of France is Paris.'",
            "timing run prompt process: \"The capital of France is 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": 1791442100.47,
            "capped_samples": 13.0,
            "cold_init_ms": 2026.65,
            "cpu_capped_samples": 13.0,
            "decode_tok_s_it1": 25.3,
            "decode_tok_s_it2": 23.91,
            "decode_tok_s_it3": 21.54,
            "decode_tokens": 256.0,
            "init_ms": 455.25,
            "iterations": 3.0,
            "max_num_tokens": 2048.0,
            "memavail_at_bench_start_kib": 4106308.0,
            "memavail_min_kib": 3357612.0,
            "prefill_tok_s_it1": 240.66,
            "prefill_tok_s_it2": 212.78,
            "prefill_tok_s_it3": 201.24,
            "prefill_tokens": 1024.0,
            "sampler_samples": 17.0,
            "t_1024_256_s": 16.2117,
            "thermal_status_at_bench_start": 1.0,
            "ttft_s_it1": 4.29,
            "ttft_s_it2": 4.85,
            "ttft_s_it3": 5.13
          },
          "output_match": null,
          "peak_mem_mb": 1534.82,
          "prefill_tokens_per_s": 207.01,
          "provenance": "measured",
          "runs": true,
          "total_ops": 818,
          "ttft_ms": 4990.0
        },
        "source": "data/device_runs/prebuilt-adac974b/2026-10-12/lfm2-700m-int8__pixel-8a.json"
      },
      {
        "device": "pixel-8a",
        "run": {
          "accelerator": "gpu",
          "context_length": null,
          "date": "2026-10-12",
          "decode_tokens_per_s": 34.75,
          "delegated_ops": 560,
          "env": {
            "device": "Pixel 8a",
            "machine_label": "pixel-8a-aab1",
            "os_build": "Android 16",
            "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 20:40:55-21:06:15 JST: [21:00:55] bundle on device: ccc95007d868878ac749073cbc28b88097cb7d437fca3f9aec72f535cc7fd99b (808281472 B, push 0 s) = Mac",
            "bench args: --backend=gpu --model_path=/data/local/tmp/aa2_700m/LFM2-700M_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/aa2_700m/metrics.pb",
            "VERBOSE: Replacing 560 out of 560 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 0 (prefill_1024).",
            "VERBOSE: Replacing 560 out of 560 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 1 (prefill_512).",
            "VERBOSE: Replacing 560 out of 560 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 2 (prefill_256).",
            "VERBOSE: Replacing 560 out of 560 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 3 (prefill_128).",
            "VERBOSE: Replacing 560 out of 560 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 4 (prefill_64).",
            "VERBOSE: Replacing 560 out of 560 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 5 (prefill_32).",
            "VERBOSE: Replacing 560 out of 560 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 6 (prefill_16).",
            "VERBOSE: Replacing 560 out of 560 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 7 (prefill_8).",
            "VERBOSE: Replacing 560 out of 560 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 8 (prefill_4).",
            "VERBOSE: Replacing 560 out of 560 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 9 (prefill_2).",
            "VERBOSE: Replacing 501 out of 501 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 10 (prefill_1).",
            "VERBOSE: Replacing 519 out of 519 node(s) with delegate (LITERT_CL) node, yielding 1 partitions for subgraph 11 (decode).",
            "[21:04:16] parsed: i8_gpu_timing gpu: gate PASS 1/1 | bench prefill 1451.0500000000002 decode 34.75 init 1877.09 cold 19417.83 peak 1107.56 capped 0/11 | answers: 'The capital of France is Paris.'",
            "timing run prompt process: \"The capital of France is Paris.\"",
            "check run (hold 2026-10-12 21:25:52-21:36:25 JST): 1 of 1 prompt processes answered with the expected word (\"The capital of France is 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": 1791442099.99,
            "capped_samples": 0.0,
            "cold_init_ms": 19417.83,
            "cpu_capped_samples": 3.0,
            "decode_tok_s_it1": 32.91,
            "decode_tok_s_it2": 34.84,
            "decode_tok_s_it3": 34.66,
            "decode_tokens": 256.0,
            "init_ms": 1877.09,
            "iterations": 3.0,
            "max_num_tokens": 2048.0,
            "memavail_at_bench_start_kib": 3977272.0,
            "memavail_min_kib": 2388324.0,
            "prefill_tok_s_it1": 1374.84,
            "prefill_tok_s_it2": 1449.71,
            "prefill_tok_s_it3": 1452.39,
            "prefill_tokens": 1024.0,
            "sampler_samples": 11.0,
            "t_1024_256_s": 8.0726,
            "thermal_status_at_bench_start": 1.0,
            "ttft_s_it1": 0.78,
            "ttft_s_it2": 0.74,
            "ttft_s_it3": 0.73
          },
          "output_match": null,
          "peak_mem_mb": 1107.56,
          "prefill_tokens_per_s": 1451.05,
          "provenance": "measured",
          "runs": true,
          "total_ops": 560,
          "ttft_ms": 735.0
        },
        "source": "data/device_runs/prebuilt-adac974b/2026-10-12/lfm2-700m-int8__pixel-8a.json"
      }
    ]
  },
  "model": {
    "family": "lfm2",
    "id": "lfm2-700m-int8",
    "license": "other (lfm1.0)",
    "source_url": "https://huggingface.co/litert-community/LFM2-700M",
    "task": "text-generation"
  },
  "pitfalls": [
    "Runtime: tested on litert-lm 0.18.0 on a Mac; on litert-lm 0.17.1 the plain `run` of this file answered as if no question had been asked (card L21).",
    "Template: Liquid AI's chat template fails at render on litert-lm 0.18.0 (message.get(\"content\")), 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|>; marked non-thinking, with the configuration's thought channel kept and no reply in the checks containing its strings (card L71, L112).",
    "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_start|> / <|tool_list_end|> and <|tool_response_start|> / <|tool_response_end|> markers, wraps the tool result in [ … ] and quotes call arguments with double quotes (card L73).",
    "One Engine per conversation: in a probe with two different system prompts the shared Engine gave the same answers and token count as fresh Engines (3 of 3 processes); LiteRT-LM issue #3165 reports recurrent state reused across conversations that share prompt tokens, and the probe's two conversations started with different system prompts; the card advises a new Engine for each conversation (card L75).",
    "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 L116).",
    "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 (no min_p or repetition penalty); --top-k 1 gives greedy output (card L69).",
    "Cache: without --cache, `run` writes a compiled-weights cache next to the model (747 MB for this file on CPU); without --backend, `run` uses the CPU (card L67).",
    "Greedy replies of this file differed between processes on the CPU path: the same file run twice on IFEval-lite flipped 5 prompts (24 prompts re-run gave byte-identical replies for 4) (card L57).",
    "Speed: int4 is faster on the Mac and Galaxy S26 GPUs, equal on the Pixel 8a GPU (8.07 s each for a 1024-token prefill plus 256 decoded tokens); on the CPUs this file is faster on the Mac (3.72 vs 4.18 s) and the Galaxy S26 (6.76 vs 7.43 s); not measured on iOS, not run on the NPU (card L30-L39, L106).",
    "Phone conditions: on the Galaxy S26 the harness waited for thermal status 0 and uncapped clocks before each process of this file's rows, and in its CPU run the CPU clock was lowered in 9 of 10 samples; on the Pixel 8a every run started at thermal status 1, and the CPU clock was lowered in 13 of 17 samples of the CPU run and 3 of 11 of the GPU run (card L97, L106).",
    "Accuracy (greedy, litert-lm on the Mac CPU, comparable only within the card's table): IFEval-lite 75 and 74 of 120 in two runs (bf16 77), GSM8K 58 and 62 of 100 with two wordings of the prompt (bf16 59 and 59); 8-question check 7/8 on CPU and GPU, the miss being the rhyme question as for bf16 (card L43, L49-L55).",
    "License: the LFM Open License v1.0's section 5 conditions commercial use on an annual revenue below 10 million US dollars (card L120)."
  ],
  "schema_version": "1.2"
}
