{
  "base_url": "http://127.0.0.1:39269/v1",
  "completed": 16,
  "concurrency": 8,
  "context_distribution": {
    "4096": 16
  },
  "context_seed": 0,
  "context_tokens": 4096,
  "engine": "sglang-5f55db35-dflash2-k16",
  "failed": 0,
  "failures": [],
  "finished_at": "2026-09-05T03:18:22Z",
  "gpu": "RTX PRO 6000 Blackwell Max-Q TP1",
  "manifest": null,
  "max_context_tokens": 262144,
  "max_tokens": 512,
  "measurement_protocol": "capacity-v3",
  "metrics": {
    "content_chunks": 231,
    "content_chunks_s": null,
    "decode_tok_s_max": 99.19510558182571,
    "decode_tok_s_mean": 72.74269211946546,
    "decode_tok_s_min": 53.725565777563794,
    "decode_tok_s_p25": 62.73079426806458,
    "decode_tok_s_p50": 71.46812914440856,
    "decode_tok_s_p75": 77.11928023602148,
    "decode_tok_s_p90": 91.71153346489314,
    "decode_tok_s_p95": 99.19510558182571,
    "decode_tok_s_p99": 99.19510558182571,
    "decode_tok_s_samples": 16,
    "e2e_max_ms": 1179.2770000174642,
    "e2e_mean_ms": 916.5272312457091,
    "e2e_min_ms": 746.8824000097811,
    "e2e_p25_ms": 807.3971000267193,
    "e2e_p50_ms": 868.3424999471754,
    "e2e_p75_ms": 1004.1753000114113,
    "e2e_p90_ms": 1092.2644999809563,
    "e2e_p95_ms": 1179.2770000174642,
    "e2e_p99_ms": 1179.2770000174642,
    "e2e_samples": 16,
    "effective_prefill_tok_s_max": 23422.406325451266,
    "effective_prefill_tok_s_mean": 16641.026130587725,
    "effective_prefill_tok_s_min": 5843.6338028049395,
    "effective_prefill_tok_s_p25": 12686.77260759146,
    "effective_prefill_tok_s_p50": 17259.548642697308,
    "effective_prefill_tok_s_p75": 17913.212400491142,
    "effective_prefill_tok_s_p90": 23355.652822634802,
    "effective_prefill_tok_s_p95": 23422.406325451266,
    "effective_prefill_tok_s_p99": 23422.406325451266,
    "effective_prefill_tok_s_samples": 16,
    "generation_max_ms": 893.4294000500813,
    "generation_mean_ms": 672.3167437667144,
    "generation_min_ms": 473.81370002403855,
    "generation_p25_ms": 545.2469000592828,
    "generation_p50_ms": 669.0287000965327,
    "generation_p75_ms": 748.8315999507904,
    "generation_p90_ms": 799.6950000524521,
    "generation_p95_ms": 893.4294000500813,
    "generation_p99_ms": 893.4294000500813,
    "generation_samples": 16,
    "mean_inter_token_latency_ms_max": 18.61311250104336,
    "mean_inter_token_latency_ms_mean": 14.089419748414073,
    "mean_inter_token_latency_ms_min": 10.081142553702948,
    "mean_inter_token_latency_ms_p25": 12.50023414698861,
    "mean_inter_token_latency_ms_p50": 13.938097918677764,
    "mean_inter_token_latency_ms_p75": 15.600658332308134,
    "mean_inter_token_latency_ms_p90": 16.66031250109275,
    "mean_inter_token_latency_ms_p95": 18.61311250104336,
    "mean_inter_token_latency_ms_p99": 18.61311250104336,
    "mean_inter_token_latency_ms_samples": 16,
    "output_token_sources": {
      "usage": 16
    },
    "output_tokens": 779,
    "prefix_cache_hit_avg": null,
    "prompt_token_samples": 16,
    "prompt_tokens": 57814,
    "prompt_tokens_max": 3614.0,
    "prompt_tokens_mean": 3613.375,
    "prompt_tokens_min": 3613.0,
    "prompt_tokens_p25": 3613.0,
    "prompt_tokens_p50": 3613.0,
    "prompt_tokens_p75": 3614.0,
    "prompt_tokens_p90": 3614.0,
    "prompt_tokens_p95": 3614.0,
    "prompt_tokens_p99": 3614.0,
    "prompt_tokens_samples": 16,
    "reasoning_chunks": 0,
    "throughput_tok_s": 342.75480795372835,
    "time_to_first_output_max_ms": 618.4507999569178,
    "time_to_first_output_mean_ms": 244.21048747899476,
    "time_to_first_output_min_ms": 154.29669991135597,
    "time_to_first_output_p25_ms": 159.05080002266914,
    "time_to_first_output_p50_ms": 208.77839997410774,
    "time_to_first_output_p75_ms": 284.5998000120744,
    "time_to_first_output_p90_ms": 285.84759996738285,
    "time_to_first_output_p95_ms": 618.4507999569178,
    "time_to_first_output_p99_ms": 618.4507999569178,
    "time_to_first_output_samples": 16,
    "tpot_max_ms": 18.61311250104336,
    "tpot_mean_ms": 14.089419748414073,
    "tpot_min_ms": 10.081142553702948,
    "tpot_p25_ms": 12.50023414698861,
    "tpot_p50_ms": 13.938097918677764,
    "tpot_p75_ms": 15.600658332308134,
    "tpot_p90_ms": 16.66031250109275,
    "tpot_p95_ms": 18.61311250104336,
    "tpot_p99_ms": 18.61311250104336,
    "tpot_samples": 16,
    "ttft_max_ms": 618.4509999584407,
    "ttft_mean_ms": 244.2107312308508,
    "ttft_min_ms": 154.29689991287887,
    "ttft_p25_ms": 159.05110002495348,
    "ttft_p50_ms": 208.77859997563064,
    "ttft_p75_ms": 284.6000000135973,
    "ttft_p90_ms": 285.84779996890575,
    "ttft_p95_ms": 618.4509999584407,
    "ttft_p99_ms": 618.4509999584407,
    "ttft_samples": 16
  },
  "model": "qwen38-27b-inferact-nvfp4-pro6000-dflash2-k16",
  "request_timings": [
    {
      "content_chunks": 17,
      "decode_tok_s": 53.725565777563794,
      "e2e_ms": 1179.2770000174642,
      "effective_prefill_tok_s": 12639.602362980371,
      "generation_ms": 893.4294000500813,
      "mean_inter_token_latency_ms": 18.61311250104336,
      "output_token_source": "usage",
      "output_tokens": 49,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3613,
      "reasoning_chunks": 0,
      "request_index": 0,
      "time_to_first_output_ms": 285.84759996738285,
      "tpot_ms": 18.61311250104336,
      "ttft_ms": 285.84779996890575,
      "visible_generation_ms": 893.4292000485584
    },
    {
      "content_chunks": 16,
      "decode_tok_s": 64.09985903793901,
      "e2e_ms": 1033.6163999745622,
      "effective_prefill_tok_s": 12686.77260759146,
      "generation_ms": 748.8315999507904,
      "mean_inter_token_latency_ms": 15.600658332308134,
      "output_token_source": "usage",
      "output_tokens": 49,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3613,
      "reasoning_chunks": 0,
      "request_index": 1,
      "time_to_first_output_ms": 284.7848000237718,
      "tpot_ms": 15.600658332308134,
      "ttft_ms": 284.78530002757907,
      "visible_generation_ms": 748.8310999469832
    },
    {
      "content_chunks": 17,
      "decode_tok_s": 60.75983977100942,
      "e2e_ms": 1074.923500069417,
      "effective_prefill_tok_s": 12680.39644807688,
      "generation_ms": 789.9955000029877,
      "mean_inter_token_latency_ms": 16.458239583395578,
      "output_token_source": "usage",
      "output_tokens": 49,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3613,
      "reasoning_chunks": 0,
      "request_index": 2,
      "time_to_first_output_ms": 284.9280000664294,
      "tpot_ms": 16.458239583395578,
      "ttft_ms": 284.9281000671908,
      "visible_generation_ms": 789.9954000022262
    },
    {
      "content_chunks": 13,
      "decode_tok_s": 86.19948136319502,
      "e2e_ms": 829.8467000713572,
      "effective_prefill_tok_s": 12695.019461878452,
      "generation_ms": 545.2469000592828,
      "mean_inter_token_latency_ms": 11.600997873601761,
      "output_token_source": "usage",
      "output_tokens": 48,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3613,
      "reasoning_chunks": 0,
      "request_index": 3,
      "time_to_first_output_ms": 284.5998000120744,
      "tpot_ms": 11.600997873601761,
      "ttft_ms": 284.6000000135973,
      "visible_generation_ms": 545.2467000577599
    },
    {
      "content_chunks": 11,
      "decode_tok_s": 79.99850148734268,
      "e2e_ms": 784.0963000198826,
      "effective_prefill_tok_s": 13303.302407991527,
      "generation_ms": 512.509600026533,
      "mean_inter_token_latency_ms": 12.50023414698861,
      "output_token_source": "usage",
      "output_tokens": 42,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3613,
      "reasoning_chunks": 0,
      "request_index": 4,
      "time_to_first_output_ms": 271.5866999933496,
      "tpot_ms": 12.50023414698861,
      "ttft_ms": 271.58719999715686,
      "visible_generation_ms": 512.5091000227258
    },
    {
      "content_chunks": 15,
      "decode_tok_s": 62.73079426806458,
      "e2e_ms": 958.566699991934,
      "effective_prefill_tok_s": 17259.548642697308,
      "generation_ms": 749.233299982734,
      "mean_inter_token_latency_ms": 15.94113404218583,
      "output_token_source": "usage",
      "output_tokens": 48,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3613,
      "reasoning_chunks": 0,
      "request_index": 5,
      "time_to_first_output_ms": 209.33340000919998,
      "tpot_ms": 15.94113404218583,
      "ttft_ms": 209.33360001072288,
      "visible_generation_ms": 749.2330999812111
    },
    {
      "content_chunks": 15,
      "decode_tok_s": 69.11215217249382,
      "e2e_ms": 896.4928999776021,
      "effective_prefill_tok_s": 17888.830800239488,
      "generation_ms": 694.5233000442386,
      "mean_inter_token_latency_ms": 14.469235417588303,
      "output_token_source": "usage",
      "output_tokens": 49,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3613,
      "reasoning_chunks": 0,
      "request_index": 6,
      "time_to_first_output_ms": 201.9695999333635,
      "tpot_ms": 14.469235417588303,
      "ttft_ms": 201.96969993412495,
      "visible_generation_ms": 694.5232000434771
    },
    {
      "content_chunks": 13,
      "decode_tok_s": 91.71153346489314,
      "e2e_ms": 746.8824000097811,
      "effective_prefill_tok_s": 17913.212400491142,
      "generation_ms": 545.1877000741661,
      "mean_inter_token_latency_ms": 10.903754001483321,
      "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": 201.69469993561506,
      "tpot_ms": 10.903754001483321,
      "ttft_ms": 201.69489993713796,
      "visible_generation_ms": 545.1875000726432
    },
    {
      "content_chunks": 15,
      "decode_tok_s": 60.02288372048303,
      "e2e_ms": 1004.1753000114113,
      "effective_prefill_tok_s": 17669.18378310848,
      "generation_ms": 799.6950000524521,
      "mean_inter_token_latency_ms": 16.66031250109275,
      "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": 204.48029995895922,
      "tpot_ms": 16.66031250109275,
      "ttft_ms": 204.48039995972067,
      "visible_generation_ms": 799.6949000516906
    },
    {
      "content_chunks": 13,
      "decode_tok_s": 77.11928023602148,
      "e2e_ms": 807.3971000267193,
      "effective_prefill_tok_s": 22716.012742375693,
      "generation_ms": 648.3463000040501,
      "mean_inter_token_latency_ms": 12.966926000081003,
      "output_token_source": "usage",
      "output_tokens": 51,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3613,
      "reasoning_chunks": 0,
      "request_index": 9,
      "time_to_first_output_ms": 159.05080002266914,
      "tpot_ms": 12.966926000081003,
      "ttft_ms": 159.05110002495348,
      "visible_generation_ms": 648.3460000017658
    },
    {
      "content_chunks": 13,
      "decode_tok_s": 74.0469006906636,
      "e2e_ms": 807.2833999758586,
      "effective_prefill_tok_s": 22723.042953622204,
      "generation_ms": 648.2377999927849,
      "mean_inter_token_latency_ms": 13.50495416651635,
      "output_token_source": "usage",
      "output_tokens": 49,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3614,
      "reasoning_chunks": 0,
      "request_index": 10,
      "time_to_first_output_ms": 159.04559998307377,
      "tpot_ms": 13.50495416651635,
      "ttft_ms": 159.04579998459667,
      "visible_generation_ms": 648.237599991262
    },
    {
      "content_chunks": 13,
      "decode_tok_s": 72.87237473718739,
      "e2e_ms": 853.7415999453515,
      "effective_prefill_tok_s": 17310.219833317053,
      "generation_ms": 644.9631999712437,
      "mean_inter_token_latency_ms": 13.72262127598391,
      "output_token_source": "usage",
      "output_tokens": 48,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3614,
      "reasoning_chunks": 0,
      "request_index": 11,
      "time_to_first_output_ms": 208.77839997410774,
      "tpot_ms": 13.72262127598391,
      "ttft_ms": 208.77859997563064,
      "visible_generation_ms": 644.9629999697208
    },
    {
      "content_chunks": 15,
      "decode_tok_s": 71.46812914440856,
      "e2e_ms": 868.3424999471754,
      "effective_prefill_tok_s": 23355.652822634802,
      "generation_ms": 713.6047999374568,
      "mean_inter_token_latency_ms": 13.99225097916582,
      "output_token_source": "usage",
      "output_tokens": 52,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3614,
      "reasoning_chunks": 0,
      "request_index": 12,
      "time_to_first_output_ms": 154.73770000971854,
      "tpot_ms": 13.99225097916582,
      "ttft_ms": 154.73790001124144,
      "visible_generation_ms": 713.604599935934
    },
    {
      "content_chunks": 14,
      "decode_tok_s": 69.07487142908153,
      "e2e_ms": 834.7177999094129,
      "effective_prefill_tok_s": 23422.406325451266,
      "generation_ms": 680.4210999980569,
      "mean_inter_token_latency_ms": 14.47704468080972,
      "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": 154.29669991135597,
      "tpot_ms": 14.47704468080972,
      "ttft_ms": 154.29689991287887,
      "visible_generation_ms": 680.420899996534
    },
    {
      "content_chunks": 16,
      "decode_tok_s": 71.7458010292745,
      "e2e_ms": 892.8116000024602,
      "effective_prefill_tok_s": 16149.580694142545,
      "generation_ms": 669.0287000965327,
      "mean_inter_token_latency_ms": 13.938097918677764,
      "output_token_source": "usage",
      "output_tokens": 49,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3614,
      "reasoning_chunks": 0,
      "request_index": 14,
      "time_to_first_output_ms": 223.78289990592748,
      "tpot_ms": 13.938097918677764,
      "ttft_ms": 223.78339990973473,
      "visible_generation_ms": 669.0282000927255
    },
    {
      "content_chunks": 15,
      "decode_tok_s": 99.19510558182571,
      "e2e_ms": 1092.2644999809563,
      "effective_prefill_tok_s": 5843.6338028049395,
      "generation_ms": 473.81370002403855,
      "mean_inter_token_latency_ms": 10.081142553702948,
      "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": 618.4507999569178,
      "tpot_ms": 10.081142553702948,
      "ttft_ms": 618.4509999584407,
      "visible_generation_ms": 473.81350002251565
    }
  ],
  "requests": 16,
  "run_id": "benchmark-20260905T031819Z",
  "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": "2026-09-05T03:18:19Z",
  "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",
    "distributions": "request-level mean, min, p25, p50, p75, p90, p95, p99, and max using the nearest-rank percentile method",
    "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",
    "tpot": "alias of mean_inter_token_latency for capacity-v3: client-observed generation time divided by completion tokens after the first token",
    "ttft": "request start through the first non-empty streamed content delta"
  },
  "wall_clock_ms": 2272.7616999763995
}
