{
  "base_url": "http://127.0.0.1:39274/v1",
  "completed": 16,
  "concurrency": 8,
  "context_distribution": {
    "4096": 16
  },
  "context_seed": 0,
  "context_tokens": 4096,
  "engine": "sglang",
  "failed": 0,
  "failures": [],
  "finished_at": "2026-09-05T03:37:42Z",
  "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": 227,
    "content_chunks_s": null,
    "decode_tok_s_max": 102.46936881752707,
    "decode_tok_s_mean": 79.75696130755374,
    "decode_tok_s_min": 52.95095682719209,
    "decode_tok_s_p25": 63.760312240053345,
    "decode_tok_s_p50": 79.32608367996457,
    "decode_tok_s_p75": 89.31936293722961,
    "decode_tok_s_p90": 99.3008804293058,
    "decode_tok_s_p95": 102.46936881752707,
    "decode_tok_s_p99": 102.46936881752707,
    "decode_tok_s_samples": 16,
    "e2e_max_ms": 1071.6629000380635,
    "e2e_mean_ms": 844.7903124761069,
    "e2e_min_ms": 611.5017998963594,
    "e2e_p25_ms": 679.8206999665126,
    "e2e_p50_ms": 863.9397999504581,
    "e2e_p75_ms": 936.0975000308827,
    "e2e_p90_ms": 1043.795199948363,
    "e2e_p95_ms": 1071.6629000380635,
    "e2e_p99_ms": 1071.6629000380635,
    "e2e_samples": 16,
    "effective_prefill_tok_s_max": 23141.668905816598,
    "effective_prefill_tok_s_mean": 17734.483353583997,
    "effective_prefill_tok_s_min": 5552.389808637538,
    "effective_prefill_tok_s_p25": 16292.628963146835,
    "effective_prefill_tok_s_p50": 18531.95903811772,
    "effective_prefill_tok_s_p75": 18802.038566471845,
    "effective_prefill_tok_s_p90": 23125.124007738246,
    "effective_prefill_tok_s_p95": 23141.668905816598,
    "effective_prefill_tok_s_p99": 23141.668905816598,
    "effective_prefill_tok_s_samples": 16,
    "generation_max_ms": 854.0408000117168,
    "generation_mean_ms": 617.7581812444259,
    "generation_min_ms": 419.56609999760985,
    "generation_p25_ms": 483.3793999860063,
    "generation_p50_ms": 630.3096999181435,
    "generation_p75_ms": 740.9690000349656,
    "generation_p90_ms": 853.827100014314,
    "generation_p95_ms": 854.0408000117168,
    "generation_p99_ms": 854.0408000117168,
    "generation_samples": 16,
    "mean_inter_token_latency_ms_max": 18.88539999878655,
    "mean_inter_token_latency_ms_mean": 13.13203067276662,
    "mean_inter_token_latency_ms_min": 9.759013952557432,
    "mean_inter_token_latency_ms_p25": 10.262731706905292,
    "mean_inter_token_latency_ms_p50": 11.439663044217488,
    "mean_inter_token_latency_ms_p75": 15.436854167395117,
    "mean_inter_token_latency_ms_p90": 17.792516666910767,
    "mean_inter_token_latency_ms_p95": 18.88539999878655,
    "mean_inter_token_latency_ms_p99": 18.88539999878655,
    "mean_inter_token_latency_ms_samples": 16,
    "output_token_sources": {
      "usage": 16
    },
    "output_tokens": 766,
    "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": 361.5942018019849,
    "time_to_first_output_max_ms": 650.8909000549465,
    "time_to_first_output_mean_ms": 227.03213123168098,
    "time_to_first_output_min_ms": 156.12529998179525,
    "time_to_first_output_p25_ms": 191.86419993638992,
    "time_to_first_output_p50_ms": 194.9581999797374,
    "time_to_first_output_p75_ms": 196.4412999805063,
    "time_to_first_output_p90_ms": 292.6594999153167,
    "time_to_first_output_p95_ms": 650.8909000549465,
    "time_to_first_output_p99_ms": 650.8909000549465,
    "time_to_first_output_samples": 16,
    "tpot_max_ms": 18.88539999878655,
    "tpot_mean_ms": 13.13203067276662,
    "tpot_min_ms": 9.759013952557432,
    "tpot_p25_ms": 10.262731706905292,
    "tpot_p50_ms": 11.439663044217488,
    "tpot_p75_ms": 15.436854167395117,
    "tpot_p90_ms": 17.792516666910767,
    "tpot_p95_ms": 18.88539999878655,
    "tpot_p99_ms": 18.88539999878655,
    "tpot_samples": 16,
    "ttft_max_ms": 650.8912000572309,
    "ttft_mean_ms": 227.03244372678455,
    "ttft_min_ms": 156.12569998484105,
    "ttft_p25_ms": 191.86439993791282,
    "ttft_p50_ms": 194.9583999812603,
    "ttft_p75_ms": 196.4414999820292,
    "ttft_p90_ms": 292.65999991912395,
    "ttft_p95_ms": 650.8912000572309,
    "ttft_p99_ms": 650.8912000572309,
    "ttft_samples": 16
  },
  "model": "qwen38-27b-inferact-nvfp4-pro6000-dflash2-k12-compile",
  "request_timings": [
    {
      "content_chunks": 13,
      "decode_tok_s": 99.3008804293058,
      "e2e_ms": 679.8206999665126,
      "effective_prefill_tok_s": 18392.26272865499,
      "generation_ms": 483.3793999860063,
      "mean_inter_token_latency_ms": 10.070404166375132,
      "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": 196.4412999805063,
      "tpot_ms": 10.070404166375132,
      "ttft_ms": 196.4414999820292,
      "visible_generation_ms": 483.3791999844834
    },
    {
      "content_chunks": 15,
      "decode_tok_s": 63.760312240053345,
      "e2e_ms": 936.0975000308827,
      "effective_prefill_tok_s": 19713.21176965616,
      "generation_ms": 752.8193999314681,
      "mean_inter_token_latency_ms": 15.68373749857225,
      "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": 183.2781000994146,
      "tpot_ms": 15.68373749857225,
      "ttft_ms": 183.2785999868065,
      "visible_generation_ms": 752.8189000440761
    },
    {
      "content_chunks": 14,
      "decode_tok_s": 87.41516215422789,
      "e2e_ms": 721.7900999821723,
      "effective_prefill_tok_s": 18474.619263089113,
      "generation_ms": 526.2245000340044,
      "mean_inter_token_latency_ms": 11.439663044217488,
      "output_token_source": "usage",
      "output_tokens": 47,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3613,
      "reasoning_chunks": 0,
      "request_index": 2,
      "time_to_first_output_ms": 195.5655999481678,
      "tpot_ms": 11.439663044217488,
      "ttft_ms": 195.56589995045215,
      "visible_generation_ms": 526.2242000317201
    },
    {
      "content_chunks": 13,
      "decode_tok_s": 89.3083806977989,
      "e2e_ms": 721.7346999095753,
      "effective_prefill_tok_s": 18483.81554319416,
      "generation_ms": 526.2664000038058,
      "mean_inter_token_latency_ms": 11.197157446889484,
      "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": 195.46829990576953,
      "tpot_ms": 11.197157446889484,
      "ttft_ms": 195.46859990805387,
      "visible_generation_ms": 526.2661000015214
    },
    {
      "content_chunks": 16,
      "decode_tok_s": 64.78003802822376,
      "e2e_ms": 935.9295000322163,
      "effective_prefill_tok_s": 18531.95903811772,
      "generation_ms": 740.9690000349656,
      "mean_inter_token_latency_ms": 15.436854167395117,
      "output_token_source": "usage",
      "output_tokens": 49,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3613,
      "reasoning_chunks": 0,
      "request_index": 4,
      "time_to_first_output_ms": 194.96049999725074,
      "tpot_ms": 15.436854167395117,
      "ttft_ms": 194.96079999953508,
      "visible_generation_ms": 740.9687000326812
    },
    {
      "content_chunks": 14,
      "decode_tok_s": 79.32608367996457,
      "e2e_ms": 825.2678998978809,
      "effective_prefill_tok_s": 18532.177668728527,
      "generation_ms": 630.3096999181435,
      "mean_inter_token_latency_ms": 12.606193998362869,
      "output_token_source": "usage",
      "output_tokens": 51,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3613,
      "reasoning_chunks": 0,
      "request_index": 5,
      "time_to_first_output_ms": 194.9581999797374,
      "tpot_ms": 12.606193998362869,
      "ttft_ms": 194.9583999812603,
      "visible_generation_ms": 630.3094999166206
    },
    {
      "content_chunks": 16,
      "decode_tok_s": 52.95095682719209,
      "e2e_ms": 1043.795199948363,
      "effective_prefill_tok_s": 18628.301199701295,
      "generation_ms": 849.8429999453947,
      "mean_inter_token_latency_ms": 18.88539999878655,
      "output_token_source": "usage",
      "output_tokens": 46,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3613,
      "reasoning_chunks": 0,
      "request_index": 6,
      "time_to_first_output_ms": 193.9522000029683,
      "tpot_ms": 18.88539999878655,
      "ttft_ms": 193.95270000677556,
      "visible_generation_ms": 849.8424999415874
    },
    {
      "content_chunks": 13,
      "decode_tok_s": 88.8552937796545,
      "e2e_ms": 677.9560999711975,
      "effective_prefill_tok_s": 18621.484444728987,
      "generation_ms": 483.9329000096768,
      "mean_inter_token_latency_ms": 11.254253488597135,
      "output_token_source": "usage",
      "output_tokens": 44,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3613,
      "reasoning_chunks": 0,
      "request_index": 7,
      "time_to_first_output_ms": 194.02319996152073,
      "tpot_ms": 11.254253488597135,
      "ttft_ms": 194.02329996228218,
      "visible_generation_ms": 483.93280000891536
    },
    {
      "content_chunks": 16,
      "decode_tok_s": 56.2034038647117,
      "e2e_ms": 1010.2778000291437,
      "effective_prefill_tok_s": 23125.124007738246,
      "generation_ms": 854.0408000117168,
      "mean_inter_token_latency_ms": 17.792516666910767,
      "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": 156.2370000174269,
      "tpot_ms": 17.792516666910767,
      "ttft_ms": 156.23730001971126,
      "visible_generation_ms": 854.0405000094324
    },
    {
      "content_chunks": 16,
      "decode_tok_s": 56.21747072585925,
      "e2e_ms": 1009.9523999961093,
      "effective_prefill_tok_s": 23141.668905816598,
      "generation_ms": 853.827100014314,
      "mean_inter_token_latency_ms": 17.78806458363154,
      "output_token_source": "usage",
      "output_tokens": 49,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3613,
      "reasoning_chunks": 0,
      "request_index": 9,
      "time_to_first_output_ms": 156.12529998179525,
      "tpot_ms": 17.78806458363154,
      "ttft_ms": 156.12569998484105,
      "visible_generation_ms": 853.8267000112683
    },
    {
      "content_chunks": 13,
      "decode_tok_s": 77.84987367995093,
      "e2e_ms": 864.3202999373898,
      "effective_prefill_tok_s": 16274.990604423198,
      "generation_ms": 642.2618000069633,
      "mean_inter_token_latency_ms": 12.845236000139266,
      "output_token_source": "usage",
      "output_tokens": 51,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3614,
      "reasoning_chunks": 0,
      "request_index": 10,
      "time_to_first_output_ms": 222.05849993042648,
      "tpot_ms": 12.845236000139266,
      "ttft_ms": 222.05879993271083,
      "visible_generation_ms": 642.261500004679
    },
    {
      "content_chunks": 14,
      "decode_tok_s": 73.19484764474534,
      "e2e_ms": 863.9397999504581,
      "effective_prefill_tok_s": 16292.628963146835,
      "generation_ms": 642.1216999879107,
      "mean_inter_token_latency_ms": 13.662163829530014,
      "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": 221.81809996254742,
      "tpot_ms": 13.662163829530014,
      "ttft_ms": 221.81839996483177,
      "visible_generation_ms": 642.1213999856263
    },
    {
      "content_chunks": 16,
      "decode_tok_s": 89.31936293722961,
      "e2e_ms": 930.8190000010654,
      "effective_prefill_tok_s": 12348.82175718109,
      "generation_ms": 638.1595000857487,
      "mean_inter_token_latency_ms": 11.19578070325875,
      "output_token_source": "usage",
      "output_tokens": 58,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3614,
      "reasoning_chunks": 0,
      "request_index": 12,
      "time_to_first_output_ms": 292.6594999153167,
      "tpot_ms": 11.19578070325875,
      "ttft_ms": 292.65999991912395,
      "visible_generation_ms": 638.1590000819415
    },
    {
      "content_chunks": 12,
      "decode_tok_s": 97.72000168801428,
      "e2e_ms": 611.7793000303209,
      "effective_prefill_tok_s": 18802.038566471845,
      "generation_ms": 419.56609999760985,
      "mean_inter_token_latency_ms": 10.233319512136825,
      "output_token_source": "usage",
      "output_tokens": 42,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3614,
      "reasoning_chunks": 0,
      "request_index": 13,
      "time_to_first_output_ms": 192.21320003271103,
      "tpot_ms": 10.233319512136825,
      "ttft_ms": 192.21350003499538,
      "visible_generation_ms": 419.5657999953255
    },
    {
      "content_chunks": 12,
      "decode_tok_s": 102.46936881752707,
      "e2e_ms": 611.5017998963594,
      "effective_prefill_tok_s": 18836.23938805767,
      "generation_ms": 419.6375999599695,
      "mean_inter_token_latency_ms": 9.759013952557432,
      "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": 191.86419993638992,
      "tpot_ms": 9.759013952557432,
      "ttft_ms": 191.86439993791282,
      "visible_generation_ms": 419.6373999584466
    },
    {
      "content_chunks": 14,
      "decode_tok_s": 97.43994372640071,
      "e2e_ms": 1071.6629000380635,
      "effective_prefill_tok_s": 5552.389808637538,
      "generation_ms": 420.771999983117,
      "mean_inter_token_latency_ms": 10.262731706905292,
      "output_token_source": "usage",
      "output_tokens": 42,
      "planned_context_tokens": 4096,
      "prompt_tokens": 3614,
      "reasoning_chunks": 0,
      "request_index": 15,
      "time_to_first_output_ms": 650.8909000549465,
      "tpot_ms": 10.262731706905292,
      "ttft_ms": 650.8912000572309,
      "visible_generation_ms": 420.77169998083264
    }
  ],
  "requests": 16,
  "run_id": "benchmark-20260905T033740Z",
  "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": "2026-09-05T03:37:40Z",
  "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": 2118.3968000113964
}
