{
  "artifacts": [
    {
      "file": "Qwen3.5-9B_int8.litertlm",
      "sha256": "82d23246c88d957ab316960147be5e04adc7b3f4492548406dcd6540fbeb92f1",
      "size_mb": 8774.598
    }
  ],
  "benchmarks": [],
  "conversion": {
    "command": "convert_qwen35_9b.sh (hf-to-litertlm; Qwen/Qwen3.5-9B at c2022362): text-decoder export (6 prefill signatures 1024..1 + decode, a 4096-position KV budget, 64 state buffers) -> int8 / Mixed INT4 quantization with the 48 in_proj_a / in_proj_b projections float -> fp32 GPU activations declared; then the tokenizer step (the repo's tokenizer.json as the tokenizer section) and use_switch_template.sh (the Qwen3.5-4B files' template, thinking switch and thought channel)",
    "quantization": "dynamic int8 on every FULLY_CONNECTED + int8 embedding; the convs, the delta rule and the 48 GatedDeltaNet decay/gate projections (in_proj_a / in_proj_b) float; fp32 GPU activations declared (card L30, L108)",
    "tool": "litert-torch (pinned commit 115a1360 + the Qwen3.5 hybrid patch of hf-to-litertlm, 2026-08-21 version; lightweight export with a 1e8-byte lazy-constant threshold) + post-hoc ai-edge-quantizer; repro = hf-to-litertlm qwen35_work/",
    "tool_version": "litert-torch 115a1360 + patch, transformers 5.14.1, torch 2.12.1, ai-edge-quantizer 0.8.0; packed and checked with litert-lm 0.18.0"
  },
  "cross_runtime": [],
  "delegation": null,
  "device": {
    "records": [
      {
        "device": "mac-studio-m4-max",
        "run": {
          "accelerator": "cpu",
          "context_length": null,
          "date": "2026-10-16",
          "decode_tokens_per_s": null,
          "delegated_ops": null,
          "env": {
            "device": "Mac Studio (M4 Max, Mac16,9)",
            "machine_label": "mac-studio-m4-max",
            "os_build": "macOS 27.0 (26A428)",
            "runtime": "litert-lm",
            "runtime_version": "0.18.0",
            "soc": "Apple M4 Max",
            "vendor_sdk": null
          },
          "error": null,
          "evidence": [
            "harness: 8 fixed questions (scripts/verify_quality.py's checks) each with an ' Answer briefly.' suffix, one `litert-lm run <file> --prompt <question> --backend cpu --cache disk --temperature 0 --seed 0` process per question (litert-lm 0.18.0 CLI), judged by a regex per question and a degeneracy rule (a reply of repeated words or an empty reply)",
            "bundle: Qwen3.5-9B_int8.litertlm sha256 82d23246c88d957ab316960147be5e04adc7b3f4492548406dcd6540fbeb92f1 = the Hugging Face LFS oid of litert-community/Qwen3.5-9B at main (upload commit 06047197), tree API read 2026-10-17 17:15 JST; read through the XNNPACK cache dir's hard link Qwen3.5-9B_int8.litertlm",
            "score 8/8 (pass mark 6), degenerate 0; rhyme item answer \"blue.\"",
            "2026-10-16 14:30:58 JST cpu step q8 done"
          ],
          "failure_class": null,
          "full_delegation": null,
          "latency_p50_ms": null,
          "loads": true,
          "max_abs_diff": null,
          "max_rel_diff": null,
          "metrics": {
            "gate_pass_mark": 6.0,
            "gate_rows_passed": 8.0,
            "gate_rows_run": 8.0
          },
          "output_match": null,
          "peak_mem_mb": null,
          "prefill_tokens_per_s": null,
          "provenance": "measured",
          "runs": true,
          "signature": "8-question check (CLI run --cache disk)",
          "total_ops": null,
          "ttft_ms": null
        },
        "source": "data/device_runs/0.18.0/2026-10-16/qwen35-9b__mac-studio-m4-max.json"
      },
      {
        "device": "mac-studio-m4-max",
        "run": {
          "accelerator": "cpu",
          "context_length": null,
          "date": "2026-10-16",
          "decode_tokens_per_s": null,
          "delegated_ops": null,
          "env": {
            "device": "Mac Studio (M4 Max, Mac16,9)",
            "machine_label": "mac-studio-m4-max",
            "os_build": "macOS 27.0 (26A428)",
            "runtime": "litert-lm",
            "runtime_version": "0.17.1",
            "soc": "Apple M4 Max",
            "vendor_sdk": null
          },
          "error": null,
          "evidence": [
            "command (tools/t1/slotpass.sh, \"$(F $n)\"): uvx --from litert-lm==0.17.1 litert-lm run \"$(F $n)\" --backend cpu --prompt \"What is the capital of France? Answer in one word.\" --cache no",
            "bundle: Qwen3.5-9B_int8.litertlm sha256 82d23246c88d957ab316960147be5e04adc7b3f4492548406dcd6540fbeb92f1 = the Hugging Face LFS oid of litert-community/Qwen3.5-9B at main (upload commit 06047197), tree API read 2026-10-17 17:15 JST",
            "RAM check: held 3 samples after 120 s (reclaimable 103.7 GB (need 70))",
            "stdout: Paris",
            "EXIT=0"
          ],
          "failure_class": null,
          "full_delegation": null,
          "latency_p50_ms": null,
          "loads": true,
          "max_abs_diff": null,
          "max_rel_diff": null,
          "metrics": {
            "answers_with_expect": 1.0,
            "processes": 1.0
          },
          "output_match": null,
          "peak_mem_mb": null,
          "prefill_tokens_per_s": null,
          "provenance": "measured",
          "runs": true,
          "signature": "one prompt (CLI run --cache no)",
          "total_ops": null,
          "ttft_ms": null
        },
        "source": "data/device_runs/0.17.1/2026-10-16/qwen35-9b__mac-studio-m4-max.json"
      },
      {
        "device": "mac-studio-m4-max",
        "run": {
          "accelerator": "cpu",
          "context_length": null,
          "date": "2026-10-16",
          "decode_tokens_per_s": null,
          "delegated_ops": null,
          "env": {
            "device": "Mac Studio (M4 Max, Mac16,9)",
            "machine_label": "mac-studio-m4-max",
            "os_build": "macOS 27.0 (26A428)",
            "runtime": "litert-lm",
            "runtime_version": "0.18.0",
            "soc": "Apple M4 Max",
            "vendor_sdk": null
          },
          "error": null,
          "evidence": [
            "harness: gate_multiturn_t1.py on the litert-lm 0.18.0 wheel's Engine / Conversation API, one process = one Engine = one Conversation, SamplerConfig(top_k=1, seed=0), max_num_tokens 4096, CPU, XNNPACK cache dir; no ThinkingConfig (the template's default = thinking off), 256 output tokens per turn: \"Give me one word that means 'very big'.\" / \"Use it in a short sentence.\" / \"Now make that sentence a question.\"; PASS = every reply non-empty with no chat marker and the token count growing",
            "bundle: Qwen3.5-9B_int8.litertlm sha256 82d23246c88d957ab316960147be5e04adc7b3f4492548406dcd6540fbeb92f1 = the Hugging Face LFS oid of litert-community/Qwen3.5-9B at main (upload commit 06047197), tree API read 2026-10-17 17:15 JST; read as",
            "turn 1: reply \"Colossal\", token_count -> 25, thought 0 characters",
            "turn 2: reply \"The **colossal** statue dominated the entire city square.\", token_count -> 57, thought 0 characters",
            "turn 3: reply \"Did the **colossal** statue dominate the entire city square?\", token_count -> 90, thought 0 characters"
          ],
          "failure_class": null,
          "full_delegation": null,
          "latency_p50_ms": null,
          "loads": true,
          "max_abs_diff": null,
          "max_rel_diff": null,
          "metrics": {
            "engine_s": 18.5,
            "max_num_tokens": 4096.0,
            "max_output_tokens": 256.0,
            "turn1_elapsed_s": 3.7,
            "turn1_kv_before_decode": 22.0,
            "turn1_render_tokens": 22.0,
            "turn1_thought_chars": 0.0,
            "turn1_token_count": 25.0,
            "turn2_elapsed_s": 1.3,
            "turn2_kv_before_decode": 44.0,
            "turn2_render_tokens": 41.0,
            "turn2_thought_chars": 0.0,
            "turn2_token_count": 57.0,
            "turn3_elapsed_s": 1.4,
            "turn3_kv_before_decode": 76.0,
            "turn3_render_tokens": 70.0,
            "turn3_thought_chars": 0.0,
            "turn3_token_count": 90.0,
            "turns_passed": 3.0
          },
          "output_match": null,
          "peak_mem_mb": null,
          "prefill_tokens_per_s": null,
          "provenance": "measured",
          "runs": true,
          "signature": "three-turn conversation, thinking off (Python API)",
          "total_ops": null,
          "ttft_ms": null
        },
        "source": "data/device_runs/0.18.0/2026-10-16/qwen35-9b__mac-studio-m4-max.json"
      },
      {
        "device": "mac-studio-m4-max",
        "run": {
          "accelerator": "cpu",
          "context_length": null,
          "date": "2026-10-16",
          "decode_tokens_per_s": null,
          "delegated_ops": null,
          "env": {
            "device": "Mac Studio (M4 Max, Mac16,9)",
            "machine_label": "mac-studio-m4-max",
            "os_build": "macOS 27.0 (26A428)",
            "runtime": "litert-lm",
            "runtime_version": "0.18.0",
            "soc": "Apple M4 Max",
            "vendor_sdk": null
          },
          "error": null,
          "evidence": [
            "harness: gate_multiturn_t1.py on the litert-lm 0.18.0 wheel's Engine / Conversation API, one process = one Engine = one Conversation, SamplerConfig(top_k=1, seed=0), max_num_tokens 4096, CPU, XNNPACK cache dir; ThinkingConfig(enable_thinking=True, thinking_token_budget=300), 2,048 output tokens per turn: \"My name is Kenji.\" / \"What is 17 times 23?\" / \"What is my name?\"; PASS = a closed thought every turn, 391 in reply 2, Kenji in reply 3, and the KV before each decode growing and equal to the token count of the bundle template's render (the runtime takes each thought out of the KV before the next turn)",
            "bundle: Qwen3.5-9B_int8.litertlm sha256 82d23246c88d957ab316960147be5e04adc7b3f4492548406dcd6540fbeb92f1 = the Hugging Face LFS oid of litert-community/Qwen3.5-9B at main (upload commit 06047197), tree API read 2026-10-17 17:15 JST; read as",
            "turn 1: reply \"Nice to meet you, Kenji! It's great to connect with you. Is there anything specific you'd like to talk about or need help with today?\", token_count -> 352, thought 1014 characters",
            "turn 2: reply \"17 times 23 is **391**.\n\nHere is the breakdown:\n*   17 × 20 = 340\n*   17 × 3 = 51\n*   340 + 51 = 391\", token_count -> 434, thought 708 characters",
            "turn 3: reply \"Your name is Kenji.\", token_count -> 455, thought 1059 characters"
          ],
          "failure_class": null,
          "full_delegation": null,
          "latency_p50_ms": null,
          "loads": true,
          "max_abs_diff": null,
          "max_rel_diff": null,
          "metrics": {
            "engine_s": 19.3,
            "max_num_tokens": 4096.0,
            "max_output_tokens": 2048.0,
            "turn1_elapsed_s": 24.9,
            "turn1_kv_before_decode": 16.0,
            "turn1_render_tokens": 16.0,
            "turn1_thought_chars": 1014.0,
            "turn1_token_count": 352.0,
            "turn2_elapsed_s": 25.1,
            "turn2_kv_before_decode": 68.0,
            "turn2_render_tokens": 68.0,
            "turn2_thought_chars": 708.0,
            "turn2_token_count": 434.0,
            "turn3_elapsed_s": 22.4,
            "turn3_kv_before_decode": 145.0,
            "turn3_render_tokens": 145.0,
            "turn3_thought_chars": 1059.0,
            "turn3_token_count": 455.0,
            "turns_passed": 3.0
          },
          "output_match": null,
          "peak_mem_mb": null,
          "prefill_tokens_per_s": null,
          "provenance": "measured",
          "runs": true,
          "signature": "three-turn conversation, thinking on with thinking_token_budget 300 (Python API)",
          "total_ops": null,
          "ttft_ms": null
        },
        "source": "data/device_runs/0.18.0/2026-10-16/qwen35-9b__mac-studio-m4-max.json"
      },
      {
        "device": "mac-studio-m4-max",
        "run": {
          "accelerator": "gpu",
          "context_length": null,
          "date": "2026-10-16",
          "decode_tokens_per_s": null,
          "delegated_ops": 34373,
          "env": {
            "device": "Mac Studio (M4 Max, Mac16,9)",
            "machine_label": "mac-studio-m4-max",
            "os_build": "macOS 27.0 (26A428)",
            "runtime": "litert-lm",
            "runtime_version": "0.18.0",
            "soc": "Apple M4 Max",
            "vendor_sdk": null
          },
          "error": null,
          "evidence": [
            "harness: 8 fixed questions (scripts/verify_quality.py's checks) each with an ' Answer briefly.' suffix, one `litert-lm run <file> --prompt <question> --backend gpu --cache no --temperature 0 --seed 0` process per question (litert-lm 0.18.0 CLI), judged by a regex per question and a degeneracy rule (a reply of repeated words or an empty reply)",
            "bundle: Qwen3.5-9B_int8.litertlm sha256 82d23246c88d957ab316960147be5e04adc7b3f4492548406dcd6540fbeb92f1 = the Hugging Face LFS oid of litert-community/Qwen3.5-9B at main (upload commit 06047197), tree API read 2026-10-17 17:15 JST; read as",
            "score 8/8 (pass mark 6), degenerate 0; rhyme item answer \"blue.\"",
            "2026-10-16 12:16:55 JST gpu step q8 done",
            "delegation: a separate verbose run of the same file, \"$CLI\" run \"$p\" --backend gpu --prompt \"What is the capital of France? Answer briefly.\" --cache no --temperature 0 --seed 0 --verbose (tools/t1/gate_t1.sh step verbose): 7 of 7 decoder signatures replaced whole on LITERT_WEBGPU, one partition each, no 'Validation error' / 'Shape mismatch' line, reply \"The capital of France is Paris.\"",
            "VERBOSE: Replacing 34373 out of 34373 node(s) with delegate (LITERT_WEBGPU) node, yielding 1 partitions for subgraph 0 (prefill_1024).",
            "VERBOSE: Replacing 25157 out of 25157 node(s) with delegate (LITERT_WEBGPU) node, yielding 1 partitions for subgraph 1 (prefill_256).",
            "VERBOSE: Replacing 22469 out of 22469 node(s) with delegate (LITERT_WEBGPU) node, yielding 1 partitions for subgraph 2 (prefill_64).",
            "VERBOSE: Replacing 22589 out of 22589 node(s) with delegate (LITERT_WEBGPU) node, yielding 1 partitions for subgraph 3 (prefill_16).",
            "VERBOSE: Replacing 22589 out of 22589 node(s) with delegate (LITERT_WEBGPU) node, yielding 1 partitions for subgraph 4 (prefill_4).",
            "VERBOSE: Replacing 2866 out of 2866 node(s) with delegate (LITERT_WEBGPU) node, yielding 1 partitions for subgraph 5 (prefill_1).",
            "VERBOSE: Replacing 2920 out of 2920 node(s) with delegate (LITERT_WEBGPU) node, yielding 1 partitions for subgraph 6 (decode)."
          ],
          "failure_class": null,
          "full_delegation": true,
          "latency_p50_ms": null,
          "loads": true,
          "max_abs_diff": null,
          "max_rel_diff": null,
          "metrics": {
            "gate_pass_mark": 6.0,
            "gate_rows_passed": 8.0,
            "gate_rows_run": 8.0,
            "verbose_signatures_replaced_whole": 7.0
          },
          "output_match": null,
          "peak_mem_mb": null,
          "prefill_tokens_per_s": null,
          "provenance": "measured",
          "runs": true,
          "signature": "8-question check (CLI run --backend gpu --cache no)",
          "total_ops": 34373,
          "ttft_ms": null
        },
        "source": "data/device_runs/0.18.0/2026-10-16/qwen35-9b__mac-studio-m4-max.json"
      },
      {
        "device": "mac-studio-m4-max",
        "run": {
          "accelerator": "gpu",
          "context_length": null,
          "date": "2026-10-16",
          "decode_tokens_per_s": null,
          "delegated_ops": null,
          "env": {
            "device": "Mac Studio (M4 Max, Mac16,9)",
            "machine_label": "mac-studio-m4-max",
            "os_build": "macOS 27.0 (26A428)",
            "runtime": "litert-lm",
            "runtime_version": "0.18.0",
            "soc": "Apple M4 Max",
            "vendor_sdk": null
          },
          "error": null,
          "evidence": [
            "harness: gate_multiturn_t1.py on the litert-lm 0.18.0 wheel's Engine / Conversation API, one process = one Engine = one Conversation, SamplerConfig(top_k=1, seed=0), max_num_tokens 4096, GPU, XNNPACK cache dir ../../../litert-compat/:nocache; no ThinkingConfig (the template's default = thinking off), 256 output tokens per turn: \"Give me one word that means 'very big'.\" / \"Use it in a short sentence.\" / \"Now make that sentence a question.\"; PASS = every reply non-empty with no chat marker and the token count growing",
            "bundle: Qwen3.5-9B_int8.litertlm sha256 82d23246c88d957ab316960147be5e04adc7b3f4492548406dcd6540fbeb92f1 = the Hugging Face LFS oid of litert-community/Qwen3.5-9B at main (upload commit 06047197), tree API read 2026-10-17 17:15 JST; read as",
            "turn 1: reply \"Colossal\", token_count -> 25, thought 0 characters",
            "turn 2: reply \"The **colossal** statue dominated the entire city square.\", token_count -> 57, thought 0 characters",
            "turn 3: reply \"Did the **colossal** statue dominate the entire city square?\", token_count -> 90, thought 0 characters"
          ],
          "failure_class": null,
          "full_delegation": null,
          "latency_p50_ms": null,
          "loads": true,
          "max_abs_diff": null,
          "max_rel_diff": null,
          "metrics": {
            "engine_s": 55.0,
            "max_num_tokens": 4096.0,
            "max_output_tokens": 256.0,
            "turn1_elapsed_s": 0.5,
            "turn1_kv_before_decode": 22.0,
            "turn1_render_tokens": 22.0,
            "turn1_thought_chars": 0.0,
            "turn1_token_count": 25.0,
            "turn2_elapsed_s": 0.7,
            "turn2_kv_before_decode": 44.0,
            "turn2_render_tokens": 41.0,
            "turn2_thought_chars": 0.0,
            "turn2_token_count": 57.0,
            "turn3_elapsed_s": 0.7,
            "turn3_kv_before_decode": 76.0,
            "turn3_render_tokens": 70.0,
            "turn3_thought_chars": 0.0,
            "turn3_token_count": 90.0,
            "turns_passed": 3.0
          },
          "output_match": null,
          "peak_mem_mb": null,
          "prefill_tokens_per_s": null,
          "provenance": "measured",
          "runs": true,
          "signature": "three-turn conversation, thinking off (Python API)",
          "total_ops": null,
          "ttft_ms": null
        },
        "source": "data/device_runs/0.18.0/2026-10-16/qwen35-9b__mac-studio-m4-max.json"
      },
      {
        "device": "mac-studio-m4-max",
        "run": {
          "accelerator": "gpu",
          "context_length": null,
          "date": "2026-10-16",
          "decode_tokens_per_s": null,
          "delegated_ops": null,
          "env": {
            "device": "Mac Studio (M4 Max, Mac16,9)",
            "machine_label": "mac-studio-m4-max",
            "os_build": "macOS 27.0 (26A428)",
            "runtime": "litert-lm",
            "runtime_version": "0.18.0",
            "soc": "Apple M4 Max",
            "vendor_sdk": null
          },
          "error": null,
          "evidence": [
            "harness: gate_multiturn_t1.py on the litert-lm 0.18.0 wheel's Engine / Conversation API, one process = one Engine = one Conversation, SamplerConfig(top_k=1, seed=0), max_num_tokens 4096, GPU, XNNPACK cache dir ../../../litert-compat/:nocache; ThinkingConfig(enable_thinking=True, thinking_token_budget=300), 2,048 output tokens per turn: \"My name is Kenji.\" / \"What is 17 times 23?\" / \"What is my name?\"; PASS = a closed thought every turn, 391 in reply 2, Kenji in reply 3, and the KV before each decode growing and equal to the token count of the bundle template's render (the runtime takes each thought out of the KV before the next turn)",
            "bundle: Qwen3.5-9B_int8.litertlm sha256 82d23246c88d957ab316960147be5e04adc7b3f4492548406dcd6540fbeb92f1 = the Hugging Face LFS oid of litert-community/Qwen3.5-9B at main (upload commit 06047197), tree API read 2026-10-17 17:15 JST; read as Qwen3.5-9B_int8.litertlm",
            "turn 1: reply \"Nice to meet you, Kenji! How can I help you today?\", token_count -> 335, thought 993 characters",
            "turn 2: reply \"17 times 23 is **391**.\", token_count -> 367, thought 658 characters",
            "turn 3: reply \"Your name is Kenji.\", token_count -> 377, thought 1050 characters"
          ],
          "failure_class": null,
          "full_delegation": null,
          "latency_p50_ms": null,
          "loads": true,
          "max_abs_diff": null,
          "max_rel_diff": null,
          "metrics": {
            "engine_s": 56.3,
            "max_num_tokens": 4096.0,
            "max_output_tokens": 2048.0,
            "turn1_elapsed_s": 8.5,
            "turn1_kv_before_decode": 16.0,
            "turn1_render_tokens": 16.0,
            "turn1_thought_chars": 993.0,
            "turn1_token_count": 335.0,
            "turn2_elapsed_s": 9.8,
            "turn2_kv_before_decode": 51.0,
            "turn2_render_tokens": 51.0,
            "turn2_thought_chars": 658.0,
            "turn2_token_count": 367.0,
            "turn3_elapsed_s": 9.1,
            "turn3_kv_before_decode": 78.0,
            "turn3_render_tokens": 78.0,
            "turn3_thought_chars": 1050.0,
            "turn3_token_count": 377.0,
            "turns_passed": 3.0
          },
          "output_match": null,
          "peak_mem_mb": null,
          "prefill_tokens_per_s": null,
          "provenance": "measured",
          "runs": true,
          "signature": "three-turn conversation, thinking on with thinking_token_budget 300 (Python API)",
          "total_ops": null,
          "ttft_ms": null
        },
        "source": "data/device_runs/0.18.0/2026-10-16/qwen35-9b__mac-studio-m4-max.json"
      }
    ]
  },
  "model": {
    "family": "qwen3.5",
    "id": "qwen35-9b",
    "license": "apache-2.0",
    "source_url": "https://huggingface.co/litert-community/Qwen3.5-9B",
    "task": "text-generation"
  },
  "pitfalls": [
    "Text only: the decoder of the multimodal release (8,953,803,264 of its 9,653,104,368 parameters); no image or video input. Only the 8 attention layers keep a growing KV cache (a 4096-token budget); the 24 linear-attention layers carry a constant-size conv and recurrent state (card L24).",
    "Runtime: tested on litert-lm 0.18.0 (Mac); on litert-lm 0.17.1 the file answered a plain `run` with \"Paris\" (Mac CPU, one run) (card L20).",
    "Thinking is off unless asked: the bundle carries the Qwen3.5-4B files' template (enable_thinking defaults to false) and declares a `thought` channel (<think>\\n / \\n</think>); with thinking on, the CPU needs a budget — without one a 2,048-token turn did not close on the CPU, while thinking_token_budget 300 closed every thought in the measured runs (card L26, L52, L77).",
    "fp32 GPU activations are declared on purpose: with the runtime's fp16 default the int8 weights answered 0 of 8 check questions (every reply a run of `!`), at 7.8 % less time on the same workload (card L46, L96).",
    "One Engine per conversation: on a shared Engine the second conversation starts from the first one's recurrent state (LiteRT-LM issue 3165); the probe broke the second conversation in 2 of 3 int8 runs (every reply a repeated `</think>`), 3 of 3 runs were right on a fresh Engine (card L83).",
    "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 L113).",
    "Tool calling: the template renders a tool list, the model writes the call in Qwen3.5's XML form as text, and the runtime (generic model type) does not parse or run it (LiteRT-LM issue 3927); a tool result sent back as a `tool` message is used in the answer (card L81).",
    "Tokenizer: the tokenizer section is the repo's tokenizer.json byte for byte; the exporter's class-written tokenizer (no `\\p{M}` in the pre-tokenizer) had tokenized Hindi, Thai, Tamil and Arabic with harakat differently; the runtime now matches the tokenizers library in 346 of 346 strings (card L79).",
    "Init: litert-lm 0.18.0's CLI prints an init time that sums overlapping init phases (LiteRT-LM #4057), about twice the engine build in the log's timestamps; the GPU build is long because each of the 7 signature graphs compiles on one thread (34,373 nodes for the 1024-token prefill) (card L87).",
    "Galaxy S26: the int8 file was not started on the phone's GPU after a sibling 8B-class int8 file restarted this phone during its GPU engine warm-up (2026-10-14); the CPU was not run (by earlier CPU peak factors the file would need 13.8 GB against the phone's 11.66 GB) (card L100).",
    "Pixel 8a: not run; no Qwen3.5 Hub file built a GPU engine on this 7.75 GB Mali phone under a memory guard (card L101). iOS not measured: the file is one 8.57 GiB TFLite section, over the ~2 GiB single-section limit (card L42).",
    "Accuracy: GSM8K n=100 (thinking off, greedy, at most 2,048 output tokens, fresh process per question, Mac GPU): bf16 97 / int8 98; the int8 file agrees with bf16 on 99 of 100 questions (card L33, L56-L64).",
    "Mac: the int8 file is the faster file on the CPU (prefill 228 vs 66 tok/s for the int4 file), the Mixed INT4 file on the GPU; without --cache, `run` writes a 7.99 GB compiled-weights cache next to the model; the benchmark and GSM8K ran on files whose TFLite section is byte-identical to this file's (the tokenizer section and metadata were replaced afterwards) (card L42, L64, L77)."
  ],
  "schema_version": "1.2"
}
