{
  "base_url": "http://127.0.0.1:39273/v1",
  "completed": 16,
  "concurrency": 8,
  "context_distribution": {
    "4096": 16
  },
  "context_seed": 0,
  "context_tokens": 4096,
  "engine": "sglang",
  "failed": 0,
  "failures": [],
  "finished_at": "2026-09-05T03:28:25Z",
  "gpu": "Primary Node / RTX PRO 6000 Blackwell Max-Q / TP=1",
  "manifest": null,
  "max_context_tokens": 262144,
  "max_tokens": 512,
  "measurement_protocol": "capacity-v3",
  "metrics": {
    "content_chunks": 218,
    "content_chunks_s": null,
    "decode_tok_s_max": 108.68371303648142,
    "decode_tok_s_mean": 83.46792282973831,
    "decode_tok_s_min": 55.84436863613385,
    "decode_tok_s_p25": 70.2015696211179,
    "decode_tok_s_p50": 83.9195169898731,
    "decode_tok_s_p75": 91.97882059087124,
    "decode_tok_s_p90": 106.39300021868944,
    "decode_tok_s_p95": 108.68371303648142,
    "decode_tok_s_p99": 108.68371303648142,
    "decode_tok_s_samples": 16,
    "e2e_max_ms": 1018.9299999037758,
    "e2e_mean_ms": 763.3771937544225,
    "e2e_min_ms": 599.2812999757007,
    "e2e_p25_ms": 649.6665000449866,
    "e2e_p50_ms": 750.6988999666646,
    "e2e_p75_ms": 829.7530999407172,
    "e2e_p90_ms": 1017.6795000443235,
    "e2e_p95_ms": 1018.9299999037758,
    "e2e_p99_ms": 1018.9299999037758,
    "e2e_samples": 16,
    "effective_prefill_tok_s_max": 24398.5099050436,
    "effective_prefill_tok_s_mean": 20519.91931883161,
    "effective_prefill_tok_s_min": 12781.256611001245,
    "effective_prefill_tok_s_p25": 19481.345561813923,
    "effective_prefill_tok_s_p50": 20453.660261416706,
    "effective_prefill_tok_s_p75": 21571.61741015845,
    "effective_prefill_tok_s_p90": 24354.757300339028,
    "effective_prefill_tok_s_p95": 24398.5099050436,
    "effective_prefill_tok_s_p99": 24398.5099050436,
    "effective_prefill_tok_s_samples": 16,
    "generation_max_ms": 841.624699998647,
    "generation_mean_ms": 583.6370500110206,
    "generation_min_ms": 432.4475000612438,
    "generation_p25_ms": 483.2529000705108,
    "generation_p50_ms": 547.0464000245556,
    "generation_p75_ms": 617.0978000154719,
    "generation_p90_ms": 841.2892000051215,
    "generation_p95_ms": 841.624699998647,
    "generation_p99_ms": 841.624699998647,
    "generation_samples": 16,
    "mean_inter_token_latency_ms_max": 17.90690851060951,
    "mean_inter_token_latency_ms_mean": 12.449190351764699,
    "mean_inter_token_latency_ms_min": 9.20101063960093,
    "mean_inter_token_latency_ms_p25": 10.094229168316815,
    "mean_inter_token_latency_ms_p50": 11.70063260740236,
    "mean_inter_token_latency_ms_p75": 14.024950000351634,
    "mean_inter_token_latency_ms_p90": 17.526858333440032,
    "mean_inter_token_latency_ms_p95": 17.90690851060951,
    "mean_inter_token_latency_ms_p99": 17.90690851060951,
    "mean_inter_token_latency_ms_samples": 16,
    "output_token_sources": {
      "usage": 16
    },
    "output_tokens": 767,
    "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": 463.9596506344194,
    "time_to_first_output_max_ms": 282.75779995601624,
    "time_to_first_output_mean_ms": 179.74014374340186,
    "time_to_first_output_min_ms": 148.12379993963987,
    "time_to_first_output_p25_ms": 160.19770002458245,
    "time_to_first_output_p50_ms": 176.43210000824183,
    "time_to_first_output_p75_ms": 177.64079989865422,
    "time_to_first_output_p90_ms": 200.61020005960017,
    "time_to_first_output_p95_ms": 282.75779995601624,
    "time_to_first_output_p99_ms": 282.75779995601624,
    "time_to_first_output_samples": 16,
    "tpot_max_ms": 17.90690851060951,
    "tpot_mean_ms": 12.449190351764699,
    "tpot_min_ms": 9.20101063960093,
    "tpot_p25_ms": 10.094229168316815,
    "tpot_p50_ms": 11.70063260740236,
    "tpot_p75_ms": 14.024950000351634,
    "tpot_p90_ms": 17.526858333440032,
    "tpot_p95_ms": 17.90690851060951,
    "tpot_p99_ms": 17.90690851060951,
    "tpot_samples": 16,
    "ttft_max_ms": 282.7580999583006,
    "ttft_mean_ms": 179.74040624540066,
    "ttft_min_ms": 148.12409994192421,
    "ttft_p25_ms": 160.19790002610534,
    "ttft_p50_ms": 176.43280001357198,
    "ttft_p75_ms": 177.64099990017712,
    "ttft_p90_ms": 200.61040006112307,
    "ttft_p95_ms": 282.7580999583006,
    "ttft_p99_ms": 282.7580999583006,
    "ttft_samples": 16
  },
  "model": "qwen38-27b-inferact-nvfp4-pro6000-dflash2-k12-chunk8k",
  "request_timings": [
    {
      "content_chunks": 13,
      "decode_tok_s": 88.98032478175698,
      "e2e_ms": 650.7415000814945,
      "effective_prefill_tok_s": 21571.61741015845,
      "generation_ms": 483.2529000705108,
      "mean_inter_token_latency_ms": 11.238439536523508,
      "output_token_source": "usage",
      "output_tokens": 44,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3613,
      "reasoning_chunks": 0,
      "request_index": 0,
      "time_to_first_output_ms": 167.48860001098365,
      "tpot_ms": 11.238439536523508,
      "ttft_ms": 167.4891000147909,
      "visible_generation_ms": 483.25240006670356
    },
    {
      "content_chunks": 16,
      "decode_tok_s": 57.05529085563893,
      "e2e_ms": 1018.9299999037758,
      "effective_prefill_tok_s": 20338.79605395411,
      "generation_ms": 841.2892000051215,
      "mean_inter_token_latency_ms": 17.526858333440032,
      "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": 177.64079989865422,
      "tpot_ms": 17.526858333440032,
      "ttft_ms": 177.64099990017712,
      "visible_generation_ms": 841.2890000035986
    },
    {
      "content_chunks": 15,
      "decode_tok_s": 71.30150196435125,
      "e2e_ms": 794.1634000744671,
      "effective_prefill_tok_s": 20404.866889989986,
      "generation_ms": 617.0978000154719,
      "mean_inter_token_latency_ms": 14.024950000351634,
      "output_token_source": "usage",
      "output_tokens": 45,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3613,
      "reasoning_chunks": 0,
      "request_index": 2,
      "time_to_first_output_ms": 177.0656000589952,
      "tpot_ms": 14.024950000351634,
      "ttft_ms": 177.0658000605181,
      "visible_generation_ms": 617.097600013949
    },
    {
      "content_chunks": 13,
      "decode_tok_s": 99.42599468778427,
      "e2e_ms": 649.6665000449866,
      "effective_prefill_tok_s": 20417.83952723826,
      "generation_ms": 472.71340002771467,
      "mean_inter_token_latency_ms": 10.057731915483291,
      "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": 176.95310001727194,
      "tpot_ms": 10.057731915483291,
      "ttft_ms": 176.95330001879483,
      "visible_generation_ms": 472.71320002619177
    },
    {
      "content_chunks": 14,
      "decode_tok_s": 91.97882059087124,
      "e2e_ms": 687.4193000840023,
      "effective_prefill_tok_s": 20478.132946505895,
      "generation_ms": 510.9872000757605,
      "mean_inter_token_latency_ms": 10.872068086718308,
      "output_token_source": "usage",
      "output_tokens": 48,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3613,
      "reasoning_chunks": 0,
      "request_index": 4,
      "time_to_first_output_ms": 176.43210000824183,
      "tpot_ms": 10.872068086718308,
      "ttft_ms": 176.43280001357198,
      "visible_generation_ms": 510.98650007043034
    },
    {
      "content_chunks": 15,
      "decode_tok_s": 82.78018047417775,
      "e2e_ms": 792.7327000070363,
      "effective_prefill_tok_s": 20453.660261416706,
      "generation_ms": 616.0895000211895,
      "mean_inter_token_latency_ms": 12.080186274925284,
      "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": 176.64319998584688,
      "tpot_ms": 12.080186274925284,
      "ttft_ms": 176.64329998660833,
      "visible_generation_ms": 616.089400020428
    },
    {
      "content_chunks": 17,
      "decode_tok_s": 55.84436863613385,
      "e2e_ms": 1017.6795000443235,
      "effective_prefill_tok_s": 20522.0192750361,
      "generation_ms": 841.624699998647,
      "mean_inter_token_latency_ms": 17.90690851060951,
      "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": 176.05480004567653,
      "tpot_ms": 17.90690851060951,
      "ttft_ms": 176.05490004643798,
      "visible_generation_ms": 841.6245999978855
    },
    {
      "content_chunks": 12,
      "decode_tok_s": 108.68371303648142,
      "e2e_ms": 608.1580000463873,
      "effective_prefill_tok_s": 20562.231627054058,
      "generation_ms": 432.4475000612438,
      "mean_inter_token_latency_ms": 9.20101063960093,
      "output_token_source": "usage",
      "output_tokens": 48,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3613,
      "reasoning_chunks": 0,
      "request_index": 7,
      "time_to_first_output_ms": 175.71049998514354,
      "tpot_ms": 9.20101063960093,
      "ttft_ms": 175.710599985905,
      "visible_generation_ms": 432.4474000604823
    },
    {
      "content_chunks": 13,
      "decode_tok_s": 70.2015696211179,
      "e2e_ms": 870.1109000248834,
      "effective_prefill_tok_s": 18010.051328031164,
      "generation_ms": 669.5006999652833,
      "mean_inter_token_latency_ms": 14.244695743942197,
      "output_token_source": "usage",
      "output_tokens": 48,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3613,
      "reasoning_chunks": 0,
      "request_index": 8,
      "time_to_first_output_ms": 200.61020005960017,
      "tpot_ms": 14.244695743942197,
      "ttft_ms": 200.61040006112307,
      "visible_generation_ms": 669.5004999637604
    },
    {
      "content_chunks": 12,
      "decode_tok_s": 70.19583443862449,
      "e2e_ms": 829.7530999407172,
      "effective_prefill_tok_s": 22553.382473316302,
      "generation_ms": 669.5553999161348,
      "mean_inter_token_latency_ms": 14.245859572683718,
      "output_token_source": "usage",
      "output_tokens": 48,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3613,
      "reasoning_chunks": 0,
      "request_index": 9,
      "time_to_first_output_ms": 160.19770002458245,
      "tpot_ms": 14.245859572683718,
      "ttft_ms": 160.19790002610534,
      "visible_generation_ms": 669.5551999146119
    },
    {
      "content_chunks": 11,
      "decode_tok_s": 72.79475485542977,
      "e2e_ms": 750.6988999666646,
      "effective_prefill_tok_s": 22587.923522189194,
      "generation_ms": 590.7018999569118,
      "mean_inter_token_latency_ms": 13.737253487370042,
      "output_token_source": "usage",
      "output_tokens": 44,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3614,
      "reasoning_chunks": 0,
      "request_index": 10,
      "time_to_first_output_ms": 159.9970000097528,
      "tpot_ms": 13.737253487370042,
      "ttft_ms": 159.9972000112757,
      "visible_generation_ms": 590.7016999553889
    },
    {
      "content_chunks": 13,
      "decode_tok_s": 91.39992512107861,
      "e2e_ms": 829.8041999805719,
      "effective_prefill_tok_s": 12781.256611001245,
      "generation_ms": 547.0464000245556,
      "mean_inter_token_latency_ms": 10.940928000491112,
      "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": 282.75779995601624,
      "tpot_ms": 10.940928000491112,
      "ttft_ms": 282.7580999583006,
      "visible_generation_ms": 547.0461000222713
    },
    {
      "content_chunks": 14,
      "decode_tok_s": 83.9195169898731,
      "e2e_ms": 758.2429999019951,
      "effective_prefill_tok_s": 19402.31840821771,
      "generation_ms": 571.9765999820083,
      "mean_inter_token_latency_ms": 11.91617916629184,
      "output_token_source": "usage",
      "output_tokens": 49,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3614,
      "reasoning_chunks": 0,
      "request_index": 12,
      "time_to_first_output_ms": 186.26639991998672,
      "tpot_ms": 11.91617916629184,
      "ttft_ms": 186.26659992150962,
      "visible_generation_ms": 571.9763999804854
    },
    {
      "content_chunks": 13,
      "decode_tok_s": 85.4654644371411,
      "e2e_ms": 723.7398999277502,
      "effective_prefill_tok_s": 19481.345561813923,
      "generation_ms": 538.2290999405086,
      "mean_inter_token_latency_ms": 11.70063260740236,
      "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": 185.51079998724163,
      "tpot_ms": 11.70063260740236,
      "ttft_ms": 185.51119999028742,
      "visible_generation_ms": 538.2286999374628
    },
    {
      "content_chunks": 14,
      "decode_tok_s": 99.06650456666294,
      "e2e_ms": 632.9129000660032,
      "effective_prefill_tok_s": 24354.757300339028,
      "generation_ms": 484.5230000792071,
      "mean_inter_token_latency_ms": 10.094229168316815,
      "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": 148.38989998679608,
      "tpot_ms": 10.094229168316815,
      "ttft_ms": 148.39019998908043,
      "visible_generation_ms": 484.5227000769228
    },
    {
      "content_chunks": 13,
      "decode_tok_s": 106.39300021868944,
      "e2e_ms": 599.2812999757007,
      "effective_prefill_tok_s": 24398.5099050436,
      "generation_ms": 451.1575000360608,
      "mean_inter_token_latency_ms": 9.3991145840846,
      "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": 148.12379993963987,
      "tpot_ms": 9.3991145840846,
      "ttft_ms": 148.12409994192421,
      "visible_generation_ms": 451.15720003377646
    }
  ],
  "requests": 16,
  "run_id": "benchmark-20260905T032823Z",
  "schema": "anvil-serving.benchmark/v1",
  "serve_flags": {
    "no_thinking": false,
    "reasoning_effort": null,
    "shared_prefix_burst": false,
    "thinking_mode": "disabled"
  },
  "source_recipe": null,
  "started_at": "2026-09-05T03:28:23Z",
  "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": 1653.160999994725
}
