{
  "base_url": "http://127.0.0.1:8001/v1",
  "completed": 12,
  "concurrency": 4,
  "context_distribution": {
    "4096": 12
  },
  "context_seed": 0,
  "context_tokens": 4096,
  "engine": "vllm-glm53-exl3-b12x-fixed-mtp5-vision",
  "failed": 0,
  "failures": [],
  "finished_at": "2026-08-29T22:29:21Z",
  "gpu": "dual-rtx-pro-6000-max-q-tp2",
  "manifest": null,
  "max_context_tokens": 262144,
  "max_tokens": 256,
  "measurement_protocol": "capacity-v4-reasoning",
  "metrics": {
    "content_chunks": 164,
    "content_chunks_s": null,
    "decode_tok_s_p50": 14.938912917737591,
    "decode_tok_s_p95": 55.48622294066555,
    "e2e_p50_ms": 4490.127400000347,
    "e2e_p95_ms": 7490.809100003389,
    "effective_prefill_tok_s_p50": 2110.383434906391,
    "effective_prefill_tok_s_p95": 2900.055470404943,
    "generation_p50_ms": 3090.074799998547,
    "generation_p95_ms": 4492.868400004227,
    "mean_inter_token_latency_ms_p50": 65.74627234039461,
    "mean_inter_token_latency_ms_p95": 97.76225588233073,
    "output_token_sources": {
      "usage": 12
    },
    "output_tokens": 547,
    "prefix_cache_hit_avg": null,
    "prompt_token_samples": 12,
    "prompt_tokens": 35616,
    "prompt_tokens_p50": 2968.0,
    "prompt_tokens_p95": 2968.0,
    "reasoning_chunks": 30,
    "throughput_tok_s": 38.9556665215275,
    "time_to_first_output_p50_ms": 1400.0526000017999,
    "time_to_first_output_p95_ms": 3958.0637000035495,
    "ttft_p50_ms": 2095.2091999934055,
    "ttft_p95_ms": 4075.8989000023575,
    "ttft_samples": 12
  },
  "model": "glm53-flash-tr3-4bpw-tp2-262k-fixed-mtp5-vision",
  "request_timings": [
    {
      "content_chunks": 9,
      "decode_tok_s": 36.231581118413324,
      "e2e_ms": 4361.026300000958,
      "effective_prefill_tok_s": 814.6191643049988,
      "generation_ms": 717.6059999983408,
      "mean_inter_token_latency_ms": 27.600230769166956,
      "output_token_source": "usage",
      "output_tokens": 27,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 0,
      "time_to_first_output_ms": 3643.4203000026173,
      "ttft_ms": 3643.4225000048173,
      "visible_generation_ms": 717.6037999961409
    },
    {
      "content_chunks": 16,
      "decode_tok_s": 10.238448114784918,
      "e2e_ms": 5516.297100002703,
      "effective_prefill_tok_s": 2900.055470404943,
      "generation_ms": 4492.868400004227,
      "mean_inter_token_latency_ms": 97.67105217400493,
      "output_token_source": "usage",
      "output_tokens": 47,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 1,
      "time_to_first_output_ms": 1023.4286999984761,
      "ttft_ms": 2315.2337999999872,
      "visible_generation_ms": 3201.0633000027156
    },
    {
      "content_chunks": 12,
      "decode_tok_s": 16.896277491200948,
      "e2e_ms": 5096.623700002965,
      "effective_prefill_tok_s": 1282.1033960244047,
      "generation_ms": 2781.677800005127,
      "mean_inter_token_latency_ms": 59.18463404266228,
      "output_token_source": "usage",
      "output_tokens": 48,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 2,
      "time_to_first_output_ms": 2314.945899997838,
      "ttft_ms": 3642.7121000015177,
      "visible_generation_ms": 1453.9116000014474
    },
    {
      "content_chunks": 15,
      "decode_tok_s": 12.73796860651267,
      "e2e_ms": 7490.809100003389,
      "effective_prefill_tok_s": 749.8616053090147,
      "generation_ms": 3532.7453999998397,
      "mean_inter_token_latency_ms": 78.50545333332977,
      "output_token_source": "usage",
      "output_tokens": 46,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 3,
      "time_to_first_output_ms": 3958.0637000035495,
      "ttft_ms": 4075.8989000023575,
      "visible_generation_ms": 3414.9102000010316
    },
    {
      "content_chunks": 13,
      "decode_tok_s": 10.228896530411765,
      "e2e_ms": 4479.457599998568,
      "effective_prefill_tok_s": 2568.494113883583,
      "generation_ms": 3323.916699999245,
      "mean_inter_token_latency_ms": 97.76225588233073,
      "output_token_source": "usage",
      "output_tokens": 35,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 0,
      "request_index": 4,
      "time_to_first_output_ms": 1155.5408999993233,
      "ttft_ms": 1155.5412999950931,
      "visible_generation_ms": 3323.916300003475
    },
    {
      "content_chunks": 12,
      "decode_tok_s": 15.209987797066304,
      "e2e_ms": 4490.127400000347,
      "effective_prefill_tok_s": 2119.920351561209,
      "generation_ms": 3090.074799998547,
      "mean_inter_token_latency_ms": 65.74627234039461,
      "output_token_source": "usage",
      "output_tokens": 48,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 5,
      "time_to_first_output_ms": 1400.0526000017999,
      "ttft_ms": 2394.406600004004,
      "visible_generation_ms": 2095.720799996343
    },
    {
      "content_chunks": 13,
      "decode_tok_s": 13.949417383570326,
      "e2e_ms": 5129.760899995745,
      "effective_prefill_tok_s": 1502.3989028049855,
      "generation_ms": 3154.253599997901,
      "mean_inter_token_latency_ms": 71.6875818181341,
      "output_token_source": "usage",
      "output_tokens": 45,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 6,
      "time_to_first_output_ms": 1975.5072999978438,
      "ttft_ms": 2095.2091999934055,
      "visible_generation_ms": 3034.551700002339
    },
    {
      "content_chunks": 13,
      "decode_tok_s": 13.013317539995501,
      "e2e_ms": 4560.722399997758,
      "effective_prefill_tok_s": 2691.5107863554763,
      "generation_ms": 3457.9959999973653,
      "mean_inter_token_latency_ms": 76.844355555497,
      "output_token_source": "usage",
      "output_tokens": 46,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 7,
      "time_to_first_output_ms": 1102.7264000003925,
      "ttft_ms": 1230.8933999956935,
      "visible_generation_ms": 3329.8290000020643
    },
    {
      "content_chunks": 17,
      "decode_tok_s": 14.938912917737591,
      "e2e_ms": 4851.003600000695,
      "effective_prefill_tok_s": 2692.297434990816,
      "generation_ms": 3748.5993999944185,
      "mean_inter_token_latency_ms": 66.93927499990033,
      "output_token_source": "usage",
      "output_tokens": 57,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 8,
      "time_to_first_output_ms": 1102.4042000062764,
      "ttft_ms": 1805.8431000026758,
      "visible_generation_ms": 3045.160499998019
    },
    {
      "content_chunks": 16,
      "decode_tok_s": 18.56353272621437,
      "e2e_ms": 4162.408099997265,
      "effective_prefill_tok_s": 2097.3987309006,
      "generation_ms": 2747.3218999948585,
      "mean_inter_token_latency_ms": 53.869056862644285,
      "output_token_source": "usage",
      "output_tokens": 52,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 9,
      "time_to_first_output_ms": 1415.0862000024063,
      "ttft_ms": 2105.6807999993907,
      "visible_generation_ms": 2056.727299997874
    },
    {
      "content_chunks": 13,
      "decode_tok_s": 26.51419160323504,
      "e2e_ms": 3103.583800002525,
      "effective_prefill_tok_s": 2110.383434906391,
      "generation_ms": 1697.2042999987025,
      "mean_inter_token_latency_ms": 37.71565111108228,
      "output_token_source": "usage",
      "output_tokens": 46,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 10,
      "time_to_first_output_ms": 1406.3795000038226,
      "ttft_ms": 1525.6416000047466,
      "visible_generation_ms": 1577.9421999977785
    },
    {
      "content_chunks": 15,
      "decode_tok_s": 55.48622294066555,
      "e2e_ms": 1987.3623000021325,
      "effective_prefill_tok_s": 2687.7723203242545,
      "generation_ms": 883.1021000005421,
      "mean_inter_token_latency_ms": 18.022491836745758,
      "output_token_source": "usage",
      "output_tokens": 50,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2968,
      "reasoning_chunks": 3,
      "request_index": 11,
      "time_to_first_output_ms": 1104.2602000015904,
      "ttft_ms": 1231.695900001796,
      "visible_generation_ms": 755.6664000003366
    }
  ],
  "requests": 12,
  "run_id": "benchmark-20260829T222907Z",
  "schema": "anvil-serving.benchmark/v1",
  "serve_flags": {
    "no_thinking": false,
    "reasoning_effort": "low",
    "shared_prefix_burst": false,
    "thinking_mode": "default"
  },
  "source_recipe": null,
  "started_at": "2026-08-29T22:29:07Z",
  "tier": null,
  "timing_methodology": {
    "clock": "client time.perf_counter",
    "decode": "usage completion tokens after the first token divided by client-observed generation time; completion tokens may include reasoning tokens",
    "effective_prefill": "usage.prompt_tokens divided by client-observed time to first output; includes queueing, scheduling, prefill, and first-token work",
    "generation": "client-observed E2E minus time to first output",
    "mean_inter_token_latency": "client-observed generation time divided by completion tokens after the first token; not a raw per-token timestamp trace",
    "time_to_first_output": "request start through the first non-empty streamed reasoning or content delta",
    "ttft": "request start through the first non-empty streamed content delta"
  },
  "wall_clock_ms": 14041.602899997088
}