{
  "arm": "num_thread sweep (full scoring rules: 1 warmup + N scored, median)",
  "box_class": "cpu-mini",
  "model_tag": "qwen3.5:9b",
  "nproc": 16,
  "runs_per_setting": 5,
  "rows": [
    {
      "num_thread": 12,
      "decode_median": 10.1,
      "decode_min": 10.093,
      "decode_max": 10.243,
      "loadavg_median": 9.81
    },
    {
      "num_thread": 16,
      "decode_median": 10.239,
      "decode_min": 9.578,
      "decode_max": 10.687,
      "loadavg_median": 15.01
    }
  ],
  "best_threads": 16,
  "detail": [
    {
      "model_id": "m1",
      "model_tag": "qwen3.5:9b",
      "box_class": "cpu-mini",
      "arm_label": "sweep t=12",
      "num_thread": 12,
      "started_utc": "2026-08-25T18:01:45Z",
      "nproc": 16,
      "num_predict": 256,
      "prompt_sha256": "90eedd0c53f9554ae3837674504fcb7090013d9a432a653352183c0f25a7ce5c",
      "loadavg_before_arm": "0.19 4.66 4.32 1/324 23845",
      "nvidia_before": "no nvidia-smi (no GPU on this box)",
      "ollama_version": "0.32.15",
      "runs": [
        {
          "num_thread": 12,
          "ttft_ms_CACHED_PREFIX": 185.13,
          "first_token_was_thinking": false,
          "wall_s": 25.1811,
          "eval_count": 256,
          "eval_duration_ns": 24993567000,
          "decode_tok_s": 10.243,
          "prompt_eval_count": 537,
          "prefill_tok_s_CACHE_ARTIFACT": 2960.88,
          "load_duration_ns": 565256,
          "streamed_chunks": 238,
          "response_chars": 1017,
          "thinking_chars": 0,
          "done_reason": "length",
          "run": 1
        },
        {
          "num_thread": 12,
          "ttft_ms_CACHED_PREFIX": 199.57,
          "first_token_was_thinking": false,
          "wall_s": 25.5459,
          "eval_count": 256,
          "eval_duration_ns": 25343909000,
          "decode_tok_s": 10.101,
          "prompt_eval_count": 537,
          "prefill_tok_s_CACHE_ARTIFACT": 2759.705,
          "load_duration_ns": 596999,
          "streamed_chunks": 238,
          "response_chars": 1017,
          "thinking_chars": 0,
          "done_reason": "length",
          "run": 2
        },
        {
          "num_thread": 12,
          "ttft_ms_CACHED_PREFIX": 199.07,
          "first_token_was_thinking": false,
          "wall_s": 25.5647,
          "eval_count": 256,
          "eval_duration_ns": 25363216000,
          "decode_tok_s": 10.093,
          "prompt_eval_count": 537,
          "prefill_tok_s_CACHE_ARTIFACT": 2759.805,
          "load_duration_ns": 648444,
          "streamed_chunks": 238,
          "response_chars": 1017,
          "thinking_chars": 0,
          "done_reason": "length",
          "run": 3
        },
        {
          "num_thread": 12,
          "ttft_ms_CACHED_PREFIX": 198.63,
          "first_token_was_thinking": false,
          "wall_s": 25.5467,
          "eval_count": 256,
          "eval_duration_ns": 25345702000,
          "decode_tok_s": 10.1,
          "prompt_eval_count": 537,
          "prefill_tok_s_CACHE_ARTIFACT": 2760.996,
          "load_duration_ns": 597647,
          "streamed_chunks": 238,
          "response_chars": 1017,
          "thinking_chars": 0,
          "done_reason": "length",
          "run": 4
        },
        {
          "num_thread": 12,
          "ttft_ms_CACHED_PREFIX": 200.19,
          "first_token_was_thinking": false,
          "wall_s": 25.5518,
          "eval_count": 256,
          "eval_duration_ns": 25349288000,
          "decode_tok_s": 10.099,
          "prompt_eval_count": 537,
          "prefill_tok_s_CACHE_ARTIFACT": 2744.403,
          "load_duration_ns": 623075,
          "streamed_chunks": 238,
          "response_chars": 1017,
          "thinking_chars": 0,
          "done_reason": "length",
          "run": 5
        }
      ],
      "errors": [],
      "quantization": "Q4_K_M",
      "param_size": "9.7B",
      "context_length": 262144,
      "think_mode": false,
      "think_note": "think:false accepted",
      "warmup_discarded": {
        "num_thread": 12,
        "ttft_ms_CACHED_PREFIX": 184.27,
        "first_token_was_thinking": false,
        "wall_s": 24.8666,
        "eval_count": 256,
        "eval_duration_ns": 24680044000,
        "decode_tok_s": 10.373,
        "prompt_eval_count": 537,
        "prefill_tok_s_CACHE_ARTIFACT": 2990.86,
        "load_duration_ns": 1366343,
        "streamed_chunks": 238,
        "response_chars": 1017,
        "thinking_chars": 0,
        "done_reason": "length"
      },
      "ps_after_load": [
        {
          "name": "qwen3.5:9b",
          "model": "qwen3.5:9b",
          "size": 6071601067,
          "digest": "6488c96fa5faab64bb65cbd30d4289e20e6130ef535a93ef9a49f42eda893ea7",
          "details": {
            "parent_model": "",
            "format": "gguf",
            "family": "qwen35",
            "families": [
              "qwen35"
            ],
            "parameter_size": "9.7B",
            "quantization_level": "Q4_K_M"
          },
          "expires_at": "2026-08-25T18:12:27.747142245Z",
          "size_vram": 0,
          "context_length": 4096
        }
      ],
      "size_vram": 0,
      "size_total": 6071601067,
      "cpu_only_receipt": true,
      "peak_runner_rss_gib": 6.64,
      "peak_runner_threads": 31,
      "loadavg_during_runs": {
        "min": 6.03,
        "median": 9.81,
        "max": 11.26,
        "mean": 9.45,
        "samples": 251,
        "nproc": 16
      },
      "contention_suspected": false,
      "nvidia_after": "no nvidia-smi (no GPU on this box)",
      "derived": {
        "decode_tok_s": {
          "min": 10.093,
          "median": 10.1,
          "max": 10.243,
          "mean": 10.127,
          "n": 5,
          "spread_pct": 1.49
        },
        "wall_s": {
          "min": 25.1811,
          "median": 25.5467,
          "max": 25.5647,
          "mean": 25.478,
          "n": 5,
          "spread_pct": 1.5
        },
        "eval_count": {
          "min": 256,
          "median": 256,
          "max": 256,
          "mean": 256.0,
          "n": 5,
          "spread_pct": 0.0
        },
        "thinking_chars_total": 0
      },
      "finished_utc": "2026-08-25T18:04:35Z"
    },
    {
      "model_id": "m1",
      "model_tag": "qwen3.5:9b",
      "box_class": "cpu-mini",
      "arm_label": "sweep t=16",
      "num_thread": 16,
      "started_utc": "2026-08-25T18:04:35Z",
      "nproc": 16,
      "num_predict": 256,
      "prompt_sha256": "90eedd0c53f9554ae3837674504fcb7090013d9a432a653352183c0f25a7ce5c",
      "loadavg_before_arm": "11.26 7.76 5.57 1/321 24183",
      "nvidia_before": "no nvidia-smi (no GPU on this box)",
      "ollama_version": "0.32.15",
      "runs": [
        {
          "num_thread": 16,
          "ttft_ms_CACHED_PREFIX": 173.5,
          "first_token_was_thinking": false,
          "wall_s": 26.8894,
          "eval_count": 256,
          "eval_duration_ns": 26712977000,
          "decode_tok_s": 9.583,
          "prompt_eval_count": 537,
          "prefill_tok_s_CACHE_ARTIFACT": 3234.667,
          "load_duration_ns": 2116908,
          "streamed_chunks": 238,
          "response_chars": 1048,
          "thinking_chars": 0,
          "done_reason": "length",
          "run": 1
        },
        {
          "num_thread": 16,
          "ttft_ms_CACHED_PREFIX": 180.08,
          "first_token_was_thinking": false,
          "wall_s": 25.1557,
          "eval_count": 256,
          "eval_duration_ns": 24974534000,
          "decode_tok_s": 10.25,
          "prompt_eval_count": 537,
          "prefill_tok_s_CACHE_ARTIFACT": 3090.47,
          "load_duration_ns": 741544,
          "streamed_chunks": 238,
          "response_chars": 1048,
          "thinking_chars": 0,
          "done_reason": "length",
          "run": 2
        },
        {
          "num_thread": 16,
          "ttft_ms_CACHED_PREFIX": 190.13,
          "first_token_was_thinking": false,
          "wall_s": 26.9223,
          "eval_count": 256,
          "eval_duration_ns": 26726910000,
          "decode_tok_s": 9.578,
          "prompt_eval_count": 537,
          "prefill_tok_s_CACHE_ARTIFACT": 2940.42,
          "load_duration_ns": 1775328,
          "streamed_chunks": 238,
          "response_chars": 1048,
          "thinking_chars": 0,
          "done_reason": "length",
          "run": 3
        },
        {
          "num_thread": 16,
          "ttft_ms_CACHED_PREFIX": 172.77,
          "first_token_was_thinking": false,
          "wall_s": 25.1778,
          "eval_count": 256,
          "eval_duration_ns": 25001562000,
          "decode_tok_s": 10.239,
          "prompt_eval_count": 537,
          "prefill_tok_s_CACHE_ARTIFACT": 3232.292,
          "load_duration_ns": 3192448,
          "streamed_chunks": 238,
          "response_chars": 1048,
          "thinking_chars": 0,
          "done_reason": "length",
          "run": 4
        },
        {
          "num_thread": 16,
          "ttft_ms_CACHED_PREFIX": 163.81,
          "first_token_was_thinking": false,
          "wall_s": 24.1207,
          "eval_count": 256,
          "eval_duration_ns": 23954602000,
          "decode_tok_s": 10.687,
          "prompt_eval_count": 537,
          "prefill_tok_s_CACHE_ARTIFACT": 3413.295,
          "load_duration_ns": 1507202,
          "streamed_chunks": 238,
          "response_chars": 1048,
          "thinking_chars": 0,
          "done_reason": "length",
          "run": 5
        }
      ],
      "errors": [],
      "quantization": "Q4_K_M",
      "param_size": "9.7B",
      "context_length": 262144,
      "think_mode": false,
      "think_note": "think:false accepted",
      "warmup_discarded": {
        "num_thread": 16,
        "ttft_ms_CACHED_PREFIX": 178.57,
        "first_token_was_thinking": false,
        "wall_s": 24.0041,
        "eval_count": 256,
        "eval_duration_ns": 23822171000,
        "decode_tok_s": 10.746,
        "prompt_eval_count": 537,
        "prefill_tok_s_CACHE_ARTIFACT": 3154.797,
        "load_duration_ns": 725788,
        "streamed_chunks": 238,
        "response_chars": 1048,
        "thinking_chars": 0,
        "done_reason": "length"
      },
      "ps_after_load": [
        {
          "name": "qwen3.5:9b",
          "model": "qwen3.5:9b",
          "size": 6071601067,
          "digest": "6488c96fa5faab64bb65cbd30d4289e20e6130ef535a93ef9a49f42eda893ea7",
          "details": {
            "parent_model": "",
            "format": "gguf",
            "family": "qwen35",
            "families": [
              "qwen35"
            ],
            "parameter_size": "9.7B",
            "quantization_level": "Q4_K_M"
          },
          "expires_at": "2026-08-25T18:15:16.128773632Z",
          "size_vram": 0,
          "context_length": 4096
        }
      ],
      "size_vram": 0,
      "size_total": 6071601067,
      "cpu_only_receipt": true,
      "peak_runner_rss_gib": 6.64,
      "peak_runner_threads": 35,
      "loadavg_during_runs": {
        "min": 12.9,
        "median": 15.01,
        "max": 15.73,
        "mean": 14.79,
        "samples": 251,
        "nproc": 16
      },
      "contention_suspected": false,
      "nvidia_after": "no nvidia-smi (no GPU on this box)",
      "derived": {
        "decode_tok_s": {
          "min": 9.578,
          "median": 10.239,
          "max": 10.687,
          "mean": 10.067,
          "n": 5,
          "spread_pct": 10.83
        },
        "wall_s": {
          "min": 24.1207,
          "median": 25.1778,
          "max": 26.9223,
          "mean": 25.653,
          "n": 5,
          "spread_pct": 11.13
        },
        "eval_count": {
          "min": 256,
          "median": 256,
          "max": 256,
          "mean": 256.0,
          "n": 5,
          "spread_pct": 0.0
        },
        "thinking_chars_total": 0
      },
      "finished_utc": "2026-08-25T18:07:24Z"
    }
  ],
  "finished_utc": "2026-08-25T18:07:24Z"
}