{
  "base_url": "http://127.0.0.1:39272/v1",
  "completed": 16,
  "concurrency": 8,
  "context_distribution": {
    "4096": 16
  },
  "context_seed": 0,
  "context_tokens": 4096,
  "engine": "sglang",
  "failed": 0,
  "failures": [],
  "finished_at": "2026-09-05T03:25:40Z",
  "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": 232,
    "content_chunks_s": null,
    "decode_tok_s_max": 99.1614109181146,
    "decode_tok_s_mean": 85.19866262430787,
    "decode_tok_s_min": 62.177686421208236,
    "decode_tok_s_p25": 74.25805455897212,
    "decode_tok_s_p50": 87.15115933601507,
    "decode_tok_s_p75": 91.56222804496345,
    "decode_tok_s_p90": 97.20398390502301,
    "decode_tok_s_p95": 99.1614109181146,
    "decode_tok_s_p99": 99.1614109181146,
    "decode_tok_s_samples": 16,
    "e2e_max_ms": 1040.841400041245,
    "e2e_mean_ms": 772.0883812726242,
    "e2e_min_ms": 615.3867000248283,
    "e2e_p25_ms": 703.1036000698805,
    "e2e_p50_ms": 749.2125999415293,
    "e2e_p75_ms": 777.0910999970511,
    "e2e_p90_ms": 1003.8215999957174,
    "e2e_p95_ms": 1040.841400041245,
    "e2e_p99_ms": 1040.841400041245,
    "e2e_samples": 16,
    "effective_prefill_tok_s_max": 25556.529867045967,
    "effective_prefill_tok_s_mean": 19219.93520133964,
    "effective_prefill_tok_s_min": 16712.610158447864,
    "effective_prefill_tok_s_p25": 16749.61137736418,
    "effective_prefill_tok_s_p50": 18111.528516829636,
    "effective_prefill_tok_s_p75": 19372.924228693795,
    "effective_prefill_tok_s_p90": 24323.218055215657,
    "effective_prefill_tok_s_p95": 25556.529867045967,
    "effective_prefill_tok_s_p99": 25556.529867045967,
    "effective_prefill_tok_s_samples": 16,
    "generation_max_ms": 842.3442000057548,
    "generation_mean_ms": 580.6190437579062,
    "generation_min_ms": 442.36870005261153,
    "generation_p25_ms": 487.26429999805987,
    "generation_p50_ms": 549.9891999643296,
    "generation_p75_ms": 561.3248998997733,
    "generation_p90_ms": 804.1469999589026,
    "generation_p95_ms": 842.3442000057548,
    "generation_p99_ms": 842.3442000057548,
    "generation_samples": 16,
    "mean_inter_token_latency_ms_max": 16.082939999178052,
    "mean_inter_token_latency_ms_mean": 11.93585063481977,
    "mean_inter_token_latency_ms_min": 10.084568086932315,
    "mean_inter_token_latency_ms_p25": 10.898527082948325,
    "mean_inter_token_latency_ms_p50": 11.159625533532273,
    "mean_inter_token_latency_ms_p75": 12.121158621900168,
    "mean_inter_token_latency_ms_p90": 14.523175862168186,
    "mean_inter_token_latency_ms_p95": 16.082939999178052,
    "mean_inter_token_latency_ms_p99": 16.082939999178052,
    "mean_inter_token_latency_ms_samples": 16,
    "output_token_sources": {
      "usage": 16
    },
    "output_tokens": 790,
    "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": 476.19880034152294,
    "time_to_first_output_max_ms": 216.24390000943094,
    "time_to_first_output_mean_ms": 191.46933751471806,
    "time_to_first_output_min_ms": 141.41199993900955,
    "time_to_first_output_p25_ms": 179.43619994912297,
    "time_to_first_output_p50_ms": 199.01420001406223,
    "time_to_first_output_p75_ms": 200.03910001832992,
    "time_to_first_output_p90_ms": 216.21199999935925,
    "time_to_first_output_p95_ms": 216.24390000943094,
    "time_to_first_output_p99_ms": 216.24390000943094,
    "time_to_first_output_samples": 16,
    "tpot_max_ms": 16.082939999178052,
    "tpot_mean_ms": 11.93585063481977,
    "tpot_min_ms": 10.084568086932315,
    "tpot_p25_ms": 10.898527082948325,
    "tpot_p50_ms": 11.159625533532273,
    "tpot_p75_ms": 12.121158621900168,
    "tpot_p90_ms": 14.523175862168186,
    "tpot_p95_ms": 16.082939999178052,
    "tpot_p99_ms": 16.082939999178052,
    "tpot_samples": 16,
    "ttft_max_ms": 216.2440000101924,
    "ttft_mean_ms": 191.46956876647891,
    "ttft_min_ms": 141.4122999412939,
    "ttft_p25_ms": 179.4364999514073,
    "ttft_p50_ms": 199.01450001634657,
    "ttft_p75_ms": 200.03930001985282,
    "ttft_p90_ms": 216.2121000001207,
    "ttft_p95_ms": 216.2440000101924,
    "ttft_p99_ms": 216.2440000101924,
    "ttft_samples": 16
  },
  "model": "qwen38-27b-inferact-nvfp4-pro6000-dflash2-k12-chunk1k",
  "request_timings": [
    {
      "content_chunks": 15,
      "decode_tok_s": 85.3120742818768,
      "e2e_ms": 750.9576000738889,
      "effective_prefill_tok_s": 18061.468981158858,
      "generation_ms": 550.918500055559,
      "mean_inter_token_latency_ms": 11.721670213948064,
      "output_token_source": "usage",
      "output_tokens": 48,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3613,
      "reasoning_chunks": 0,
      "request_index": 0,
      "time_to_first_output_ms": 200.03910001832992,
      "tpot_ms": 11.721670213948064,
      "ttft_ms": 200.03930001985282,
      "visible_generation_ms": 550.9183000540361
    },
    {
      "content_chunks": 14,
      "decode_tok_s": 91.75551818966302,
      "e2e_ms": 709.6267000306398,
      "effective_prefill_tok_s": 19372.924228693795,
      "generation_ms": 523.1292999815196,
      "mean_inter_token_latency_ms": 10.898527082948325,
      "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": 186.4974000491202,
      "tpot_ms": 10.898527082948325,
      "ttft_ms": 186.49790005292743,
      "visible_generation_ms": 523.1287999777123
    },
    {
      "content_chunks": 15,
      "decode_tok_s": 90.77978746859428,
      "e2e_ms": 750.2811000449583,
      "effective_prefill_tok_s": 18110.484484464945,
      "generation_ms": 550.7834000745788,
      "mean_inter_token_latency_ms": 11.015668001491576,
      "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": 199.49769997037947,
      "tpot_ms": 11.015668001491576,
      "ttft_ms": 199.49789997190237,
      "visible_generation_ms": 550.7832000730559
    },
    {
      "content_chunks": 17,
      "decode_tok_s": 62.177686421208236,
      "e2e_ms": 1003.8215999957174,
      "effective_prefill_tok_s": 18094.439649979802,
      "generation_ms": 804.1469999589026,
      "mean_inter_token_latency_ms": 16.082939999178052,
      "output_token_source": "usage",
      "output_tokens": 51,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3613,
      "reasoning_chunks": 0,
      "request_index": 3,
      "time_to_first_output_ms": 199.6746000368148,
      "tpot_ms": 16.082939999178052,
      "ttft_ms": 199.67470003757626,
      "visible_generation_ms": 804.1468999581411
    },
    {
      "content_chunks": 14,
      "decode_tok_s": 90.91087607400804,
      "e2e_ms": 749.4753999635577,
      "effective_prefill_tok_s": 18111.528516829636,
      "generation_ms": 549.9891999643296,
      "mean_inter_token_latency_ms": 10.999783999286592,
      "output_token_source": "usage",
      "output_tokens": 51,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3613,
      "reasoning_chunks": 0,
      "request_index": 4,
      "time_to_first_output_ms": 199.48619999922812,
      "tpot_ms": 10.999783999286592,
      "ttft_ms": 199.48629999998957,
      "visible_generation_ms": 549.9890999635682
    },
    {
      "content_chunks": 14,
      "decode_tok_s": 86.27665825697201,
      "e2e_ms": 709.0015000430867,
      "effective_prefill_tok_s": 18154.48344763694,
      "generation_ms": 509.9873000290245,
      "mean_inter_token_latency_ms": 11.590620455205102,
      "output_token_source": "usage",
      "output_tokens": 45,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3613,
      "reasoning_chunks": 0,
      "request_index": 5,
      "time_to_first_output_ms": 199.01420001406223,
      "tpot_ms": 11.590620455205102,
      "ttft_ms": 199.01450001634657,
      "visible_generation_ms": 509.98700002674013
    },
    {
      "content_chunks": 15,
      "decode_tok_s": 87.15115933601507,
      "e2e_ms": 749.2125999415293,
      "effective_prefill_tok_s": 18206.51927520873,
      "generation_ms": 550.767199951224,
      "mean_inter_token_latency_ms": 11.4743166656505,
      "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": 198.4453999903053,
      "tpot_ms": 11.4743166656505,
      "ttft_ms": 198.445999994874,
      "visible_generation_ms": 550.7665999466553
    },
    {
      "content_chunks": 18,
      "decode_tok_s": 68.85546312256172,
      "e2e_ms": 1040.841400041245,
      "effective_prefill_tok_s": 18201.768082139282,
      "generation_ms": 842.3442000057548,
      "mean_inter_token_latency_ms": 14.523175862168186,
      "output_token_source": "usage",
      "output_tokens": 59,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3613,
      "reasoning_chunks": 0,
      "request_index": 7,
      "time_to_first_output_ms": 198.4972000354901,
      "tpot_ms": 14.523175862168186,
      "ttft_ms": 198.49730003625154,
      "visible_generation_ms": 842.3441000049934
    },
    {
      "content_chunks": 16,
      "decode_tok_s": 82.50036413129914,
      "e2e_ms": 851.9340000348166,
      "effective_prefill_tok_s": 24263.49905349362,
      "generation_ms": 703.0272000702098,
      "mean_inter_token_latency_ms": 12.121158621900168,
      "output_token_source": "usage",
      "output_tokens": 59,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3613,
      "reasoning_chunks": 0,
      "request_index": 8,
      "time_to_first_output_ms": 148.90679996460676,
      "tpot_ms": 12.121158621900168,
      "ttft_ms": 148.90699996612966,
      "visible_generation_ms": 703.0270000686869
    },
    {
      "content_chunks": 18,
      "decode_tok_s": 74.25805455897212,
      "e2e_ms": 916.1347000626847,
      "effective_prefill_tok_s": 24323.218055215657,
      "generation_ms": 767.5934999715537,
      "mean_inter_token_latency_ms": 13.466552631079889,
      "output_token_source": "usage",
      "output_tokens": 58,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3613,
      "reasoning_chunks": 0,
      "request_index": 9,
      "time_to_first_output_ms": 148.54120009113103,
      "tpot_ms": 13.466552631079889,
      "ttft_ms": 148.54140009265393,
      "visible_generation_ms": 767.5932999700308
    },
    {
      "content_chunks": 10,
      "decode_tok_s": 91.56222804496345,
      "e2e_ms": 664.0267999609932,
      "effective_prefill_tok_s": 16712.610158447864,
      "generation_ms": 447.7828999515623,
      "mean_inter_token_latency_ms": 10.921534145160056,
      "output_token_source": "usage",
      "output_tokens": 42,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3614,
      "reasoning_chunks": 0,
      "request_index": 10,
      "time_to_first_output_ms": 216.24390000943094,
      "tpot_ms": 10.921534145160056,
      "ttft_ms": 216.2440000101924,
      "visible_generation_ms": 447.78279995080084
    },
    {
      "content_chunks": 13,
      "decode_tok_s": 89.60874152947295,
      "e2e_ms": 740.714400075376,
      "effective_prefill_tok_s": 16715.075944030443,
      "generation_ms": 524.5024000760168,
      "mean_inter_token_latency_ms": 11.159625533532273,
      "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": 216.21199999935925,
      "tpot_ms": 11.159625533532273,
      "ttft_ms": 216.2121000001207,
      "visible_generation_ms": 524.5023000752553
    },
    {
      "content_chunks": 12,
      "decode_tok_s": 94.40461778173191,
      "e2e_ms": 703.1036000698805,
      "effective_prefill_tok_s": 16743.93865620135,
      "generation_ms": 487.26429999805987,
      "mean_inter_token_latency_ms": 10.592702173870867,
      "output_token_source": "usage",
      "output_tokens": 47,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3614,
      "reasoning_chunks": 0,
      "request_index": 12,
      "time_to_first_output_ms": 215.83930007182062,
      "tpot_ms": 10.592702173870867,
      "ttft_ms": 215.83950007334352,
      "visible_generation_ms": 487.26409999653697
    },
    {
      "content_chunks": 14,
      "decode_tok_s": 71.25997796844956,
      "e2e_ms": 777.0910999970511,
      "effective_prefill_tok_s": 16749.61137736418,
      "generation_ms": 561.3248998997733,
      "mean_inter_token_latency_ms": 14.033122497494332,
      "output_token_source": "usage",
      "output_tokens": 41,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3614,
      "reasoning_chunks": 0,
      "request_index": 13,
      "time_to_first_output_ms": 215.76620009727776,
      "tpot_ms": 14.033122497494332,
      "ttft_ms": 215.76640009880066,
      "visible_generation_ms": 561.3246998982504
    },
    {
      "content_chunks": 13,
      "decode_tok_s": 97.20398390502301,
      "e2e_ms": 621.8049000017345,
      "effective_prefill_tok_s": 20140.863443523143,
      "generation_ms": 442.36870005261153,
      "mean_inter_token_latency_ms": 10.287644187270036,
      "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": 179.43619994912297,
      "tpot_ms": 10.287644187270036,
      "ttft_ms": 179.4364999514073,
      "visible_generation_ms": 442.3684000503272
    },
    {
      "content_chunks": 14,
      "decode_tok_s": 99.1614109181146,
      "e2e_ms": 615.3867000248283,
      "effective_prefill_tok_s": 25556.529867045967,
      "generation_ms": 473.97470008581877,
      "mean_inter_token_latency_ms": 10.084568086932315,
      "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": 141.41199993900955,
      "tpot_ms": 10.084568086932315,
      "ttft_ms": 141.4122999412939,
      "visible_generation_ms": 473.9744000835344
    }
  ],
  "requests": 16,
  "run_id": "benchmark-20260905T032538Z",
  "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": "2026-09-05T03:25:38Z",
  "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": 1658.9709999971092
}
