{
  "base_url": "http://127.0.0.1:39267/v1",
  "completed": 16,
  "concurrency": 8,
  "context_distribution": {
    "4096": 16
  },
  "context_seed": 0,
  "context_tokens": 4096,
  "engine": "sglang-5f55db35-dflash2-k4",
  "failed": 0,
  "failures": [],
  "finished_at": "2026-09-05T03:09:29Z",
  "gpu": "RTX PRO 6000 Blackwell Max-Q TP1",
  "manifest": null,
  "max_context_tokens": 262144,
  "max_tokens": 512,
  "measurement_protocol": "capacity-v3",
  "metrics": {
    "content_chunks": 277,
    "content_chunks_s": null,
    "decode_tok_s_p50": 77.66894026584221,
    "decode_tok_s_p95": 85.41733113702324,
    "e2e_p50_ms": 879.1045999387279,
    "e2e_p95_ms": 1286.000999971293,
    "effective_prefill_tok_s_p50": 16229.100650547793,
    "effective_prefill_tok_s_p95": 24647.020435675116,
    "generation_p50_ms": 605.5327999638394,
    "generation_p95_ms": 754.7953999601305,
    "mean_inter_token_latency_ms_p50": 12.869951062578153,
    "mean_inter_token_latency_ms_p95": 16.61822926991324,
    "output_token_sources": {
      "usage": 16
    },
    "output_tokens": 764,
    "prefix_cache_hit_avg": null,
    "prompt_token_samples": 16,
    "prompt_tokens": 57814,
    "prompt_tokens_p50": 3613.0,
    "prompt_tokens_p95": 3614.0,
    "reasoning_chunks": 0,
    "throughput_tok_s": 330.90334490811387,
    "time_to_first_output_p50_ms": 222.66500000841916,
    "time_to_first_output_p95_ms": 700.0692000146955,
    "ttft_p50_ms": 222.66520000994205,
    "ttft_p95_ms": 700.0694000162184,
    "ttft_samples": 16
  },
  "model": "qwen38-27b-inferact-nvfp4-pro6000-dflash2-k4",
  "request_timings": [
    {
      "content_chunks": 18,
      "decode_tok_s": 78.39856198148894,
      "e2e_ms": 918.3325000340119,
      "effective_prefill_tok_s": 12877.55416873213,
      "generation_ms": 637.7668000059202,
      "mean_inter_token_latency_ms": 12.755336000118405,
      "output_token_source": "usage",
      "output_tokens": 51,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3613,
      "reasoning_chunks": 0,
      "request_index": 0,
      "time_to_first_output_ms": 280.56570002809167,
      "ttft_ms": 280.56590002961457,
      "visible_generation_ms": 637.7666000043973
    },
    {
      "content_chunks": 17,
      "decode_tok_s": 77.66894026584221,
      "e2e_ms": 884.8015000112355,
      "effective_prefill_tok_s": 12918.84334916166,
      "generation_ms": 605.1325000589713,
      "mean_inter_token_latency_ms": 12.875159575722794,
      "output_token_source": "usage",
      "output_tokens": 48,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3613,
      "reasoning_chunks": 0,
      "request_index": 1,
      "time_to_first_output_ms": 279.66899995226413,
      "ttft_ms": 279.6696999575943,
      "visible_generation_ms": 605.1318000536412
    },
    {
      "content_chunks": 19,
      "decode_tok_s": 67.42271805887924,
      "e2e_ms": 1021.1468000197783,
      "effective_prefill_tok_s": 12924.019071611201,
      "generation_ms": 741.5897999890149,
      "mean_inter_token_latency_ms": 14.831795999780297,
      "output_token_source": "usage",
      "output_tokens": 51,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3613,
      "reasoning_chunks": 0,
      "request_index": 2,
      "time_to_first_output_ms": 279.55700003076345,
      "ttft_ms": 279.55720003228635,
      "visible_generation_ms": 741.589599987492
    },
    {
      "content_chunks": 16,
      "decode_tok_s": 68.43711733308503,
      "e2e_ms": 849.4811001000926,
      "effective_prefill_tok_s": 12921.342889533831,
      "generation_ms": 569.8662000941113,
      "mean_inter_token_latency_ms": 14.611953848566955,
      "output_token_source": "usage",
      "output_tokens": 40,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3613,
      "reasoning_chunks": 0,
      "request_index": 3,
      "time_to_first_output_ms": 279.6149000059813,
      "ttft_ms": 279.6151000075042,
      "visible_generation_ms": 569.8660000925884
    },
    {
      "content_chunks": 18,
      "decode_tok_s": 75.23321119462767,
      "e2e_ms": 917.3950999975204,
      "effective_prefill_tok_s": 12932.253321007083,
      "generation_ms": 638.0161000415683,
      "mean_inter_token_latency_ms": 13.29200208419934,
      "output_token_source": "usage",
      "output_tokens": 49,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3613,
      "reasoning_chunks": 0,
      "request_index": 4,
      "time_to_first_output_ms": 279.37899995595217,
      "ttft_ms": 279.37919995747507,
      "visible_generation_ms": 638.0159000400454
    },
    {
      "content_chunks": 18,
      "decode_tok_s": 79.26903382090354,
      "e2e_ms": 879.1045999387279,
      "effective_prefill_tok_s": 13206.77058209816,
      "generation_ms": 605.5327999638394,
      "mean_inter_token_latency_ms": 12.61526666591332,
      "output_token_source": "usage",
      "output_tokens": 49,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3613,
      "reasoning_chunks": 0,
      "request_index": 5,
      "time_to_first_output_ms": 273.57179997488856,
      "ttft_ms": 273.57199997641146,
      "visible_generation_ms": 605.5325999623165
    },
    {
      "content_chunks": 17,
      "decode_tok_s": 77.70037315119957,
      "e2e_ms": 792.6345999585465,
      "effective_prefill_tok_s": 19243.992841776184,
      "generation_ms": 604.8876999411732,
      "mean_inter_token_latency_ms": 12.869951062578153,
      "output_token_source": "usage",
      "output_tokens": 48,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3613,
      "reasoning_chunks": 0,
      "request_index": 6,
      "time_to_first_output_ms": 187.74690001737326,
      "ttft_ms": 187.7470000181347,
      "visible_generation_ms": 604.8875999404117
    },
    {
      "content_chunks": 17,
      "decode_tok_s": 81.14239402145816,
      "e2e_ms": 790.2579000219703,
      "effective_prefill_tok_s": 20757.54407542112,
      "generation_ms": 616.2007000530139,
      "mean_inter_token_latency_ms": 12.324014001060277,
      "output_token_source": "usage",
      "output_tokens": 51,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3613,
      "reasoning_chunks": 0,
      "request_index": 7,
      "time_to_first_output_ms": 174.0571999689564,
      "ttft_ms": 174.0577999735251,
      "visible_generation_ms": 616.2001000484452
    },
    {
      "content_chunks": 18,
      "decode_tok_s": 63.59339233192921,
      "e2e_ms": 936.9559999322519,
      "effective_prefill_tok_s": 19834.146355210447,
      "generation_ms": 754.7953999601305,
      "mean_inter_token_latency_ms": 15.724904165836051,
      "output_token_source": "usage",
      "output_tokens": 49,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3613,
      "reasoning_chunks": 0,
      "request_index": 8,
      "time_to_first_output_ms": 182.16059997212142,
      "ttft_ms": 182.16099997516721,
      "visible_generation_ms": 754.7949999570847
    },
    {
      "content_chunks": 18,
      "decode_tok_s": 63.614437134281026,
      "e2e_ms": 902.0953000290319,
      "effective_prefill_tok_s": 24486.681090399245,
      "generation_ms": 754.5457000378519,
      "mean_inter_token_latency_ms": 15.719702084121915,
      "output_token_source": "usage",
      "output_tokens": 49,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3613,
      "reasoning_chunks": 0,
      "request_index": 9,
      "time_to_first_output_ms": 147.54959999118,
      "ttft_ms": 147.54969999194145,
      "visible_generation_ms": 754.5456000370905
    },
    {
      "content_chunks": 16,
      "decode_tok_s": 85.41733113702324,
      "e2e_ms": 732.3439000174403,
      "effective_prefill_tok_s": 24587.944426510316,
      "generation_ms": 585.3613000363111,
      "mean_inter_token_latency_ms": 11.707226000726223,
      "output_token_source": "usage",
      "output_tokens": 51,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3614,
      "reasoning_chunks": 0,
      "request_index": 10,
      "time_to_first_output_ms": 146.98259998112917,
      "ttft_ms": 146.98289998341352,
      "visible_generation_ms": 585.3610000340268
    },
    {
      "content_chunks": 15,
      "decode_tok_s": 78.08728996873754,
      "e2e_ms": 697.5749999983236,
      "effective_prefill_tok_s": 24600.22925653513,
      "generation_ms": 550.6657999940217,
      "mean_inter_token_latency_ms": 12.806181395209807,
      "output_token_source": "usage",
      "output_tokens": 44,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3614,
      "reasoning_chunks": 0,
      "request_index": 11,
      "time_to_first_output_ms": 146.90920000430197,
      "ttft_ms": 146.90920000430197,
      "visible_generation_ms": 550.6657999940217
    },
    {
      "content_chunks": 16,
      "decode_tok_s": 60.17488288060072,
      "e2e_ms": 827.9777000425383,
      "effective_prefill_tok_s": 24647.020435675116,
      "generation_ms": 681.3474000664428,
      "mean_inter_token_latency_ms": 16.61822926991324,
      "output_token_source": "usage",
      "output_tokens": 42,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3614,
      "reasoning_chunks": 0,
      "request_index": 12,
      "time_to_first_output_ms": 146.63029997609556,
      "ttft_ms": 146.63069997914135,
      "visible_generation_ms": 681.347000063397
    },
    {
      "content_chunks": 19,
      "decode_tok_s": 69.25680230647407,
      "e2e_ms": 901.3201000634581,
      "effective_prefill_tok_s": 16229.100650547793,
      "generation_ms": 678.6337000085041,
      "mean_inter_token_latency_ms": 14.439014893797959,
      "output_token_source": "usage",
      "output_tokens": 48,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3614,
      "reasoning_chunks": 0,
      "request_index": 13,
      "time_to_first_output_ms": 222.686400054954,
      "ttft_ms": 222.6866000564769,
      "visible_generation_ms": 678.6335000069812
    },
    {
      "content_chunks": 16,
      "decode_tok_s": 78.6060938680088,
      "e2e_ms": 795.139699941501,
      "effective_prefill_tok_s": 16230.660408521104,
      "generation_ms": 572.4746999330819,
      "mean_inter_token_latency_ms": 12.72165999851293,
      "output_token_source": "usage",
      "output_tokens": 46,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3614,
      "reasoning_chunks": 0,
      "request_index": 14,
      "time_to_first_output_ms": 222.66500000841916,
      "ttft_ms": 222.66520000994205,
      "visible_generation_ms": 572.474499931559
    },
    {
      "content_chunks": 19,
      "decode_tok_s": 80.2141136621728,
      "e2e_ms": 1286.000999971293,
      "effective_prefill_tok_s": 5162.346807892901,
      "generation_ms": 585.9317999565974,
      "mean_inter_token_latency_ms": 12.466634041629732,
      "output_token_source": "usage",
      "output_tokens": 48,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3614,
      "reasoning_chunks": 0,
      "request_index": 15,
      "time_to_first_output_ms": 700.0692000146955,
      "ttft_ms": 700.0694000162184,
      "visible_generation_ms": 585.9315999550745
    }
  ],
  "requests": 16,
  "run_id": "benchmark-20260905T030926Z",
  "schema": "anvil-serving.benchmark/v1",
  "serve_flags": {
    "no_thinking": true,
    "reasoning_effort": null,
    "shared_prefix_burst": false,
    "thinking_mode": "disabled"
  },
  "source_recipe": null,
  "started_at": "2026-09-05T03:09:26Z",
  "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": 2308.8313000043854
}
