{
  "base_url": "http://127.0.0.1:39276/v1",
  "completed": 16,
  "concurrency": 8,
  "context_distribution": {
    "4096": 16
  },
  "context_seed": 0,
  "context_tokens": 4096,
  "engine": "sglang",
  "failed": 0,
  "failures": [],
  "finished_at": "2026-09-05T03:46:18Z",
  "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": 98.38116926755268,
    "decode_tok_s_mean": 84.72604174632143,
    "decode_tok_s_min": 62.14681961714381,
    "decode_tok_s_p25": 73.59704660750094,
    "decode_tok_s_p50": 86.81150562557852,
    "decode_tok_s_p75": 90.80861752459265,
    "decode_tok_s_p90": 96.00076443149777,
    "decode_tok_s_p95": 98.38116926755268,
    "decode_tok_s_p99": 98.38116926755268,
    "decode_tok_s_samples": 16,
    "e2e_max_ms": 1025.6780999479815,
    "e2e_mean_ms": 767.2491687626461,
    "e2e_min_ms": 621.0167999379337,
    "e2e_p25_ms": 695.6442999653518,
    "e2e_p50_ms": 737.2703000437468,
    "e2e_p75_ms": 779.3677999870852,
    "e2e_p90_ms": 989.2817999934778,
    "e2e_p95_ms": 1025.6780999479815,
    "e2e_p99_ms": 1025.6780999479815,
    "e2e_samples": 16,
    "effective_prefill_tok_s_max": 25222.793200371394,
    "effective_prefill_tok_s_mean": 19986.727437594785,
    "effective_prefill_tok_s_min": 16850.89660843933,
    "effective_prefill_tok_s_p25": 16902.22525324197,
    "effective_prefill_tok_s_p50": 19595.368920224522,
    "effective_prefill_tok_s_p75": 19960.58694670038,
    "effective_prefill_tok_s_p90": 24531.38803588113,
    "effective_prefill_tok_s_p95": 25222.793200371394,
    "effective_prefill_tok_s_p99": 25222.793200371394,
    "effective_prefill_tok_s_samples": 16,
    "generation_max_ms": 842.3144000116736,
    "generation_mean_ms": 583.5536124941427,
    "generation_min_ms": 447.91310001164675,
    "generation_p25_ms": 491.18969996925443,
    "generation_p50_ms": 552.4876000126824,
    "generation_p75_ms": 565.3507999377325,
    "generation_p90_ms": 804.5464000897482,
    "generation_p95_ms": 842.3144000116736,
    "generation_p99_ms": 842.3144000116736,
    "generation_samples": 16,
    "mean_inter_token_latency_ms_max": 16.090928001794964,
    "mean_inter_token_latency_ms_mean": 11.997720985836965,
    "mean_inter_token_latency_ms_min": 10.164546807534359,
    "mean_inter_token_latency_ms_p25": 10.903320833070515,
    "mean_inter_token_latency_ms_p50": 11.251842551072068,
    "mean_inter_token_latency_ms_p75": 12.19388620791443,
    "mean_inter_token_latency_ms_p90": 14.522662069166786,
    "mean_inter_token_latency_ms_p95": 16.090928001794964,
    "mean_inter_token_latency_ms_p99": 16.090928001794964,
    "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": 478.84713975433425,
    "time_to_first_output_max_ms": 214.46930000092834,
    "time_to_first_output_mean_ms": 183.69555626850342,
    "time_to_first_output_min_ms": 143.28309998381883,
    "time_to_first_output_p25_ms": 173.20900002960116,
    "time_to_first_output_p50_ms": 184.2254000948742,
    "time_to_first_output_p75_ms": 185.59540004935116,
    "time_to_first_output_p90_ms": 214.2354000825435,
    "time_to_first_output_p95_ms": 214.46930000092834,
    "time_to_first_output_p99_ms": 214.46930000092834,
    "time_to_first_output_samples": 16,
    "tpot_max_ms": 16.090928001794964,
    "tpot_mean_ms": 11.997720985836965,
    "tpot_min_ms": 10.164546807534359,
    "tpot_p25_ms": 10.903320833070515,
    "tpot_p50_ms": 11.251842551072068,
    "tpot_p75_ms": 12.19388620791443,
    "tpot_p90_ms": 14.522662069166786,
    "tpot_p95_ms": 16.090928001794964,
    "tpot_p99_ms": 16.090928001794964,
    "tpot_samples": 16,
    "ttft_max_ms": 214.46950000245124,
    "ttft_mean_ms": 183.69586877088295,
    "ttft_min_ms": 143.28339998610318,
    "ttft_p25_ms": 173.2097000349313,
    "ttft_p50_ms": 184.22580009792,
    "ttft_p75_ms": 185.5957000516355,
    "ttft_p90_ms": 214.2356000840664,
    "ttft_p95_ms": 214.46950000245124,
    "ttft_p99_ms": 214.46950000245124,
    "ttft_samples": 16
  },
  "model": "qwen38-27b-inferact-nvfp4-pro6000-dflash2-k12-chunk1k-mamba40",
  "request_timings": [
    {
      "content_chunks": 15,
      "decode_tok_s": 85.04276837747712,
      "e2e_ms": 738.2585000013933,
      "effective_prefill_tok_s": 19467.077303851696,
      "generation_ms": 552.6630999520421,
      "mean_inter_token_latency_ms": 11.758789360681748,
      "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": 185.59540004935116,
      "tpot_ms": 11.758789360681748,
      "ttft_ms": 185.5957000516355,
      "visible_generation_ms": 552.6627999497578
    },
    {
      "content_chunks": 14,
      "decode_tok_s": 91.71517699148428,
      "e2e_ms": 696.5684000169858,
      "effective_prefill_tok_s": 20859.19322542445,
      "generation_ms": 523.3593999873847,
      "mean_inter_token_latency_ms": 10.903320833070515,
      "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": 173.20900002960116,
      "tpot_ms": 10.903320833070515,
      "ttft_ms": 173.2097000349313,
      "visible_generation_ms": 523.3586999820545
    },
    {
      "content_chunks": 15,
      "decode_tok_s": 90.49976868051382,
      "e2e_ms": 737.2936001047492,
      "effective_prefill_tok_s": 19550.231043364784,
      "generation_ms": 552.4876000126824,
      "mean_inter_token_latency_ms": 11.049752000253648,
      "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": 184.80600009206682,
      "tpot_ms": 11.049752000253648,
      "ttft_ms": 184.80620009358972,
      "visible_generation_ms": 552.4874000111595
    },
    {
      "content_chunks": 17,
      "decode_tok_s": 62.14681961714381,
      "e2e_ms": 989.2817999934778,
      "effective_prefill_tok_s": 19557.702540405513,
      "generation_ms": 804.5464000897482,
      "mean_inter_token_latency_ms": 16.090928001794964,
      "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": 184.73539990372956,
      "tpot_ms": 16.090928001794964,
      "ttft_ms": 184.73559990525246,
      "visible_generation_ms": 804.5462000882253
    },
    {
      "content_chunks": 14,
      "decode_tok_s": 90.40857262154005,
      "e2e_ms": 737.2703000437468,
      "effective_prefill_tok_s": 19611.845044925085,
      "generation_ms": 553.0448999488726,
      "mean_inter_token_latency_ms": 11.060897998977453,
      "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": 184.2254000948742,
      "tpot_ms": 11.060897998977453,
      "ttft_ms": 184.22580009792,
      "visible_generation_ms": 553.0444999458268
    },
    {
      "content_chunks": 14,
      "decode_tok_s": 86.06121299846554,
      "e2e_ms": 695.6442999653518,
      "effective_prefill_tok_s": 19595.368920224522,
      "generation_ms": 511.26399997156113,
      "mean_inter_token_latency_ms": 11.619636362990025,
      "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": 184.3802999937907,
      "tpot_ms": 11.619636362990025,
      "ttft_ms": 184.38059999607503,
      "visible_generation_ms": 511.2636999692768
    },
    {
      "content_chunks": 15,
      "decode_tok_s": 86.81150562557852,
      "e2e_ms": 736.0235000960529,
      "effective_prefill_tok_s": 19732.2357956804,
      "generation_ms": 552.9221000615507,
      "mean_inter_token_latency_ms": 11.519210417948974,
      "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": 183.10140003450215,
      "tpot_ms": 11.519210417948974,
      "ttft_ms": 183.1017000367865,
      "visible_generation_ms": 552.9218000592664
    },
    {
      "content_chunks": 18,
      "decode_tok_s": 68.8578991397941,
      "e2e_ms": 1025.6780999479815,
      "effective_prefill_tok_s": 19704.009033712715,
      "generation_ms": 842.3144000116736,
      "mean_inter_token_latency_ms": 14.522662069166786,
      "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": 183.36369993630797,
      "tpot_ms": 14.522662069166786,
      "ttft_ms": 183.36399993859231,
      "visible_generation_ms": 842.3141000093892
    },
    {
      "content_chunks": 16,
      "decode_tok_s": 82.00830998004156,
      "e2e_ms": 854.7974000684917,
      "effective_prefill_tok_s": 24486.282800426205,
      "generation_ms": 707.2454000590369,
      "mean_inter_token_latency_ms": 12.19388620791443,
      "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": 147.5520000094548,
      "tpot_ms": 12.19388620791443,
      "ttft_ms": 147.55230001173913,
      "visible_generation_ms": 707.2451000567526
    },
    {
      "content_chunks": 18,
      "decode_tok_s": 73.59704660750094,
      "e2e_ms": 921.7683000024408,
      "effective_prefill_tok_s": 24531.38803588113,
      "generation_ms": 774.4876000797376,
      "mean_inter_token_latency_ms": 13.58750175578487,
      "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": 147.2806999227032,
      "tpot_ms": 13.58750175578487,
      "ttft_ms": 147.28119992651045,
      "visible_generation_ms": 774.4871000759304
    },
    {
      "content_chunks": 10,
      "decode_tok_s": 90.80861752459265,
      "e2e_ms": 665.9682999597862,
      "effective_prefill_tok_s": 16850.89660843933,
      "generation_ms": 451.4989999588579,
      "mean_inter_token_latency_ms": 11.01217073070385,
      "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": 214.46930000092834,
      "tpot_ms": 11.01217073070385,
      "ttft_ms": 214.46950000245124,
      "visible_generation_ms": 451.498799957335
    },
    {
      "content_chunks": 13,
      "decode_tok_s": 88.87433284468779,
      "e2e_ms": 743.0719999829307,
      "effective_prefill_tok_s": 16869.294237122107,
      "generation_ms": 528.8365999003872,
      "mean_inter_token_latency_ms": 11.251842551072068,
      "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": 214.2354000825435,
      "tpot_ms": 11.251842551072068,
      "ttft_ms": 214.2356000840664,
      "visible_generation_ms": 528.8363998988643
    },
    {
      "content_chunks": 12,
      "decode_tok_s": 93.65017223056454,
      "e2e_ms": 705.0077000167221,
      "effective_prefill_tok_s": 16902.22525324197,
      "generation_ms": 491.18969996925443,
      "mean_inter_token_latency_ms": 10.678036955853358,
      "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": 213.81800004746765,
      "tpot_ms": 10.678036955853358,
      "ttft_ms": 213.81820004899055,
      "visible_generation_ms": 491.18949996773154
    },
    {
      "content_chunks": 14,
      "decode_tok_s": 70.75253100270767,
      "e2e_ms": 779.3677999870852,
      "effective_prefill_tok_s": 16886.50901174489,
      "generation_ms": 565.3507999377325,
      "mean_inter_token_latency_ms": 14.133769998443313,
      "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": 214.01700004935265,
      "tpot_ms": 14.133769998443313,
      "ttft_ms": 214.01720005087554,
      "visible_generation_ms": 565.3505999362096
    },
    {
      "content_chunks": 13,
      "decode_tok_s": 96.00076443149777,
      "e2e_ms": 628.9699000772089,
      "effective_prefill_tok_s": 19960.58694670038,
      "generation_ms": 447.91310001164675,
      "mean_inter_token_latency_ms": 10.416583721201087,
      "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": 181.05680006556213,
      "tpot_ms": 10.416583721201087,
      "ttft_ms": 181.05720006860793,
      "visible_generation_ms": 447.91270000860095
    },
    {
      "content_chunks": 14,
      "decode_tok_s": 98.38116926755268,
      "e2e_ms": 621.0167999379337,
      "effective_prefill_tok_s": 25222.793200371394,
      "generation_ms": 477.73369995411485,
      "mean_inter_token_latency_ms": 10.164546807534359,
      "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": 143.28309998381883,
      "tpot_ms": 10.164546807534359,
      "ttft_ms": 143.28339998610318,
      "visible_generation_ms": 477.7333999518305
    }
  ],
  "requests": 16,
  "run_id": "benchmark-20260905T034617Z",
  "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": "2026-09-05T03:46:17Z",
  "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": 1649.7957999818027
}
