{
  "base_url": "http://127.0.0.1:39268/v1",
  "completed": 16,
  "concurrency": 8,
  "context_distribution": {
    "4096": 16
  },
  "context_seed": 0,
  "context_tokens": 4096,
  "engine": "sglang-5f55db35-dflash2-k12",
  "failed": 0,
  "failures": [],
  "finished_at": "2026-09-05T03:14:14Z",
  "gpu": "RTX PRO 6000 Blackwell Max-Q TP1",
  "manifest": null,
  "max_context_tokens": 262144,
  "max_tokens": 512,
  "measurement_protocol": "capacity-v3",
  "metrics": {
    "content_chunks": 236,
    "content_chunks_s": null,
    "decode_tok_s_p50": 76.34823116848249,
    "decode_tok_s_p95": 99.39121821351618,
    "e2e_p50_ms": 859.3070999486372,
    "e2e_p95_ms": 1052.2565999999642,
    "effective_prefill_tok_s_p50": 16668.481294350033,
    "effective_prefill_tok_s_p95": 20413.778771282257,
    "generation_p50_ms": 592.9254000075161,
    "generation_p95_ms": 875.2683000639081,
    "mean_inter_token_latency_ms_p50": 12.991043999791145,
    "mean_inter_token_latency_ms_p95": 17.162123530664864,
    "output_token_sources": {
      "usage": 16
    },
    "output_tokens": 781,
    "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": 421.669356066813,
    "time_to_first_output_p50_ms": 213.9471999835223,
    "time_to_first_output_p95_ms": 321.42060005571693,
    "ttft_p50_ms": 213.94749998580664,
    "ttft_p95_ms": 321.42080005723983,
    "ttft_samples": 16
  },
  "model": "qwen38-27b-inferact-nvfp4-pro6000-dflash2-k12",
  "request_timings": [
    {
      "content_chunks": 15,
      "decode_tok_s": 71.79165344403978,
      "e2e_ms": 923.5868999967352,
      "effective_prefill_tok_s": 14169.433163254102,
      "generation_ms": 668.6013999860734,
      "mean_inter_token_latency_ms": 13.929195833043195,
      "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": 254.98550001066178,
      "ttft_ms": 254.98580001294613,
      "visible_generation_ms": 668.601099983789
    },
    {
      "content_chunks": 15,
      "decode_tok_s": 70.6241289710721,
      "e2e_ms": 923.1388999614865,
      "effective_prefill_tok_s": 14838.726901656793,
      "generation_ms": 679.6543999807909,
      "mean_inter_token_latency_ms": 14.159466666266477,
      "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": 243.48449998069555,
      "ttft_ms": 243.4853999875486,
      "visible_generation_ms": 679.6534999739379
    },
    {
      "content_chunks": 16,
      "decode_tok_s": 71.57364499978894,
      "e2e_ms": 1036.7951999651268,
      "effective_prefill_tok_s": 14202.920543379096,
      "generation_ms": 782.4108999921009,
      "mean_inter_token_latency_ms": 13.971623214144659,
      "output_token_source": "usage",
      "output_tokens": 57,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3613,
      "reasoning_chunks": 0,
      "request_index": 2,
      "time_to_first_output_ms": 254.38429997302592,
      "ttft_ms": 254.38449997454882,
      "visible_generation_ms": 782.410699990578
    },
    {
      "content_chunks": 14,
      "decode_tok_s": 91.777426072946,
      "e2e_ms": 766.3763000164181,
      "effective_prefill_tok_s": 14209.428011548238,
      "generation_ms": 512.1084999991581,
      "mean_inter_token_latency_ms": 10.89592553189698,
      "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": 254.26780001726002,
      "ttft_ms": 254.2680000187829,
      "visible_generation_ms": 512.1082999976352
    },
    {
      "content_chunks": 14,
      "decode_tok_s": 79.8824710933447,
      "e2e_ms": 742.9462999571115,
      "effective_prefill_tok_s": 18804.28090943837,
      "generation_ms": 550.8092000382021,
      "mean_inter_token_latency_ms": 12.51839090995914,
      "output_token_source": "usage",
      "output_tokens": 45,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3613,
      "reasoning_chunks": 0,
      "request_index": 4,
      "time_to_first_output_ms": 192.13709991890937,
      "ttft_ms": 192.13759992271662,
      "visible_generation_ms": 550.8087000343949
    },
    {
      "content_chunks": 16,
      "decode_tok_s": 76.34823116848249,
      "e2e_ms": 859.3070999486372,
      "effective_prefill_tok_s": 18885.065070633165,
      "generation_ms": 667.9919000016525,
      "mean_inter_token_latency_ms": 13.097880392189266,
      "output_token_source": "usage",
      "output_tokens": 52,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3613,
      "reasoning_chunks": 0,
      "request_index": 5,
      "time_to_first_output_ms": 191.31519994698465,
      "ttft_ms": 191.31539994850755,
      "visible_generation_ms": 667.9917000001296
    },
    {
      "content_chunks": 13,
      "decode_tok_s": 99.39121821351618,
      "e2e_ms": 663.6255000485107,
      "effective_prefill_tok_s": 18941.349964145324,
      "generation_ms": 472.87880000658333,
      "mean_inter_token_latency_ms": 10.061251063969857,
      "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": 190.7467000419274,
      "ttft_ms": 190.7469000434503,
      "visible_generation_ms": 472.87860000506043
    },
    {
      "content_chunks": 16,
      "decode_tok_s": 61.32987705628081,
      "e2e_ms": 973.3177999733016,
      "effective_prefill_tok_s": 18949.466341300176,
      "generation_ms": 782.6527999714017,
      "mean_inter_token_latency_ms": 16.305266666070867,
      "output_token_source": "usage",
      "output_tokens": 49,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3613,
      "reasoning_chunks": 0,
      "request_index": 7,
      "time_to_first_output_ms": 190.6650000018999,
      "ttft_ms": 190.66510000266135,
      "visible_generation_ms": 782.6526999706402
    },
    {
      "content_chunks": 10,
      "decode_tok_s": 72.04860604923013,
      "e2e_ms": 799.6962000615895,
      "effective_prefill_tok_s": 16668.481294350033,
      "generation_ms": 582.9397999914363,
      "mean_inter_token_latency_ms": 13.87951904741515,
      "output_token_source": "usage",
      "output_tokens": 43,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3613,
      "reasoning_chunks": 0,
      "request_index": 8,
      "time_to_first_output_ms": 216.75640007015318,
      "ttft_ms": 216.75660007167608,
      "visible_generation_ms": 582.9395999899134
    },
    {
      "content_chunks": 17,
      "decode_tok_s": 58.26784769455973,
      "e2e_ms": 1052.2565999999642,
      "effective_prefill_tok_s": 20413.778771282257,
      "generation_ms": 875.2683000639081,
      "mean_inter_token_latency_ms": 17.162123530664864,
      "output_token_source": "usage",
      "output_tokens": 52,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3613,
      "reasoning_chunks": 0,
      "request_index": 9,
      "time_to_first_output_ms": 176.98829993605614,
      "ttft_ms": 176.98869993910193,
      "visible_generation_ms": 875.2679000608623
    },
    {
      "content_chunks": 14,
      "decode_tok_s": 77.88209463024306,
      "e2e_ms": 912.057000095956,
      "effective_prefill_tok_s": 11243.834400699669,
      "generation_ms": 590.6364000402391,
      "mean_inter_token_latency_ms": 12.839921740005197,
      "output_token_source": "usage",
      "output_tokens": 47,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3614,
      "reasoning_chunks": 0,
      "request_index": 10,
      "time_to_first_output_ms": 321.42060005571693,
      "ttft_ms": 321.42080005723983,
      "visible_generation_ms": 590.6362000387162
    },
    {
      "content_chunks": 16,
      "decode_tok_s": 73.24840717588917,
      "e2e_ms": 896.5558999916539,
      "effective_prefill_tok_s": 16892.0182188799,
      "generation_ms": 682.6087000081316,
      "mean_inter_token_latency_ms": 13.652174000162631,
      "output_token_source": "usage",
      "output_tokens": 51,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3614,
      "reasoning_chunks": 0,
      "request_index": 11,
      "time_to_first_output_ms": 213.9471999835223,
      "ttft_ms": 213.94749998580664,
      "visible_generation_ms": 682.6084000058472
    },
    {
      "content_chunks": 16,
      "decode_tok_s": 76.97610754117042,
      "e2e_ms": 863.3627999806777,
      "effective_prefill_tok_s": 16902.810244908767,
      "generation_ms": 649.5521999895573,
      "mean_inter_token_latency_ms": 12.991043999791145,
      "output_token_source": "usage",
      "output_tokens": 51,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3614,
      "reasoning_chunks": 0,
      "request_index": 12,
      "time_to_first_output_ms": 213.8105999911204,
      "ttft_ms": 213.8107999926433,
      "visible_generation_ms": 649.5519999880344
    },
    {
      "content_chunks": 14,
      "decode_tok_s": 79.19833381599707,
      "e2e_ms": 794.2875999724492,
      "effective_prefill_tok_s": 16929.993488816934,
      "generation_ms": 580.820299917832,
      "mean_inter_token_latency_ms": 12.626528259083305,
      "output_token_source": "usage",
      "output_tokens": 47,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3614,
      "reasoning_chunks": 0,
      "request_index": 13,
      "time_to_first_output_ms": 213.46730005461723,
      "ttft_ms": 213.46760005690157,
      "visible_generation_ms": 580.8199999155477
    },
    {
      "content_chunks": 13,
      "decode_tok_s": 93.52650834500555,
      "e2e_ms": 680.2527999971062,
      "effective_prefill_tok_s": 16390.758588935365,
      "generation_ms": 459.76270001847297,
      "mean_inter_token_latency_ms": 10.692155814383092,
      "output_token_source": "usage",
      "output_tokens": 44,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3614,
      "reasoning_chunks": 0,
      "request_index": 14,
      "time_to_first_output_ms": 220.49009997863322,
      "ttft_ms": 220.49039998091757,
      "visible_generation_ms": 459.7624000161886
    },
    {
      "content_chunks": 17,
      "decode_tok_s": 80.95453492023033,
      "e2e_ms": 813.4038000134751,
      "effective_prefill_tok_s": 16391.628385829736,
      "generation_ms": 592.9254000075161,
      "mean_inter_token_latency_ms": 12.352612500156587,
      "output_token_source": "usage",
      "output_tokens": 49,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3614,
      "reasoning_chunks": 0,
      "request_index": 15,
      "time_to_first_output_ms": 220.47840000595897,
      "ttft_ms": 220.47860000748187,
      "visible_generation_ms": 592.9252000059932
    }
  ],
  "requests": 16,
  "run_id": "benchmark-20260905T031412Z",
  "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": "2026-09-05T03:14:12Z",
  "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": 1852.1620999090374
}
