{
  "base_url": "redacted://routed-capability",
  "clock_domain_id": null,
  "completed": 20,
  "concurrency": 4,
  "configuration_fingerprint": null,
  "context_distribution": {
    "8192": 20
  },
  "context_seed": 941114,
  "context_tokens": 8192,
  "controlled_output_policy": "observe",
  "engine": "vllm-exl3-v84",
  "failed": 0,
  "failures": [],
  "finished_at": "2026-10-17T00:20:30Z",
  "finished_at_unix_ns": 1792196430155170706,
  "gpu": "dual-rtx-pro-6000-maxq",
  "manifest": null,
  "max_context_tokens": null,
  "max_tokens": 16384,
  "measurement_protocol": "capacity-v4-reasoning",
  "metric_population": {
    "aggregate_throughput_eligible": true,
    "attempted_requests": 20,
    "excluded_requests": 0,
    "failure_policy": "failed or contract-invalid responses retain sanitized timing and metadata but are excluded from performance distributions",
    "p99_interpretation": "descriptive-nearest-rank-only",
    "p99_interpretation_by_metric": {
      "decode_tok_s": "descriptive-nearest-rank-only",
      "e2e": "descriptive-nearest-rank-only",
      "effective_prefill_tok_s": "descriptive-nearest-rank-only",
      "tpot": "descriptive-nearest-rank-only",
      "ttft": "descriptive-nearest-rank-only"
    },
    "p99_minimum_recommended_samples": 100,
    "p99_samples_by_metric": {
      "decode_tok_s": 20,
      "e2e": 20,
      "effective_prefill_tok_s": 20,
      "tpot": 20,
      "ttft": 20
    },
    "performance_eligible_requests": 20,
    "successful_requests": 20
  },
  "metrics": {
    "completion_tokens_max": 506.0,
    "completion_tokens_mean": 239.35,
    "completion_tokens_mean_ci95_high": 274.81271094514557,
    "completion_tokens_mean_ci95_low": 203.88728905485442,
    "completion_tokens_min": 155.0,
    "completion_tokens_p25": 165.0,
    "completion_tokens_p50": 218.0,
    "completion_tokens_p75": 268.0,
    "completion_tokens_p90": 305.0,
    "completion_tokens_p95": 321.0,
    "completion_tokens_p99": 506.0,
    "completion_tokens_samples": 20,
    "completion_tokens_stddev": 80.91533912221563,
    "content_chunks": 1078,
    "content_chunks_s": null,
    "controlled_output_exact_adherence": 0,
    "controlled_output_exact_adherence_rate": null,
    "controlled_output_samples": 0,
    "controlled_output_unobservable": 0,
    "decode_tok_s_max": 56.05850326199245,
    "decode_tok_s_mean": 37.101471871626686,
    "decode_tok_s_mean_ci95_high": 39.75275538122812,
    "decode_tok_s_mean_ci95_low": 34.45018836202525,
    "decode_tok_s_min": 27.172930902371235,
    "decode_tok_s_p25": 33.295157566507406,
    "decode_tok_s_p50": 36.33154852272061,
    "decode_tok_s_p75": 38.4099405524398,
    "decode_tok_s_p90": 42.96463718119075,
    "decode_tok_s_p95": 44.2153793306485,
    "decode_tok_s_p99": 56.05850326199245,
    "decode_tok_s_samples": 20,
    "decode_tok_s_stddev": 6.04943893376839,
    "e2e_max_ms": 18510.20039105788,
    "e2e_mean_ci95_high_ms": 12318.610347551028,
    "e2e_mean_ci95_low_ms": 9800.752743846071,
    "e2e_mean_ms": 11059.68154569855,
    "e2e_min_ms": 6906.332648941316,
    "e2e_p25_ms": 8870.98974303808,
    "e2e_p50_ms": 10173.091002972797,
    "e2e_p75_ms": 12107.699107029475,
    "e2e_p90_ms": 14305.661260033958,
    "e2e_p95_ms": 16667.783229029737,
    "e2e_p99_ms": 18510.20039105788,
    "e2e_samples": 20,
    "e2e_stddev_ms": 2872.500387525006,
    "effective_prefill_tok_s_max": 2350.7026073181232,
    "effective_prefill_tok_s_mean": 1506.4861843113613,
    "effective_prefill_tok_s_mean_ci95_high": 1709.7258383805395,
    "effective_prefill_tok_s_mean_ci95_low": 1303.246530242183,
    "effective_prefill_tok_s_min": 590.0469154999577,
    "effective_prefill_tok_s_p25": 1205.139475601241,
    "effective_prefill_tok_s_p50": 1387.43986018916,
    "effective_prefill_tok_s_p75": 1731.0112098326158,
    "effective_prefill_tok_s_p90": 2103.258472760191,
    "effective_prefill_tok_s_p95": 2263.352805800933,
    "effective_prefill_tok_s_p99": 2350.7026073181232,
    "effective_prefill_tok_s_samples": 20,
    "effective_prefill_tok_s_stddev": 463.73232879818784,
    "generation_max_ms": 15299.198077991605,
    "generation_mean_ci95_high_ms": 7660.947810944485,
    "generation_mean_ci95_low_ms": 5463.341233048272,
    "generation_mean_ms": 6562.144521996379,
    "generation_min_ms": 3870.956007973291,
    "generation_p25_ms": 4563.830086961389,
    "generation_p50_ms": 6385.550002101809,
    "generation_p75_ms": 7231.834145961329,
    "generation_p90_ms": 8061.405526008457,
    "generation_p95_ms": 8321.632406092249,
    "generation_p99_ms": 15299.198077991605,
    "generation_samples": 20,
    "generation_stddev_ms": 2507.141681620724,
    "mean_inter_token_latency_ms_max": 36.80133010284641,
    "mean_inter_token_latency_ms_mean": 27.56004186331028,
    "mean_inter_token_latency_ms_mean_ci95_high": 29.33627144312785,
    "mean_inter_token_latency_ms_mean_ci95_low": 25.783812283492708,
    "mean_inter_token_latency_ms_min": 17.83850694918567,
    "mean_inter_token_latency_ms_p25": 26.005101269038278,
    "mean_inter_token_latency_ms_p50": 26.517781335554133,
    "mean_inter_token_latency_ms_p75": 29.562731491212077,
    "mean_inter_token_latency_ms_p90": 31.982172222342342,
    "mean_inter_token_latency_ms_p95": 33.26490181946537,
    "mean_inter_token_latency_ms_p99": 36.80133010284641,
    "mean_inter_token_latency_ms_samples": 20,
    "mean_inter_token_latency_ms_stddev": 4.052826616446913,
    "mean_time_per_output_token_max_ms": 36.80133010284641,
    "mean_time_per_output_token_mean_ci95_high_ms": 29.33627144312785,
    "mean_time_per_output_token_mean_ci95_low_ms": 25.783812283492708,
    "mean_time_per_output_token_mean_ms": 27.56004186331028,
    "mean_time_per_output_token_min_ms": 17.83850694918567,
    "mean_time_per_output_token_p25_ms": 26.005101269038278,
    "mean_time_per_output_token_p50_ms": 26.517781335554133,
    "mean_time_per_output_token_p75_ms": 29.562731491212077,
    "mean_time_per_output_token_p90_ms": 31.982172222342342,
    "mean_time_per_output_token_p95_ms": 33.26490181946537,
    "mean_time_per_output_token_p99_ms": 36.80133010284641,
    "mean_time_per_output_token_samples": 20,
    "mean_time_per_output_token_stddev_ms": 4.052826616446913,
    "output_token_sources": {
      "usage": 20
    },
    "output_tokens": 4787,
    "prefix_cache_hit_avg": null,
    "prompt_token_samples": 20,
    "prompt_tokens": 121453,
    "prompt_tokens_max": 6079.0,
    "prompt_tokens_mean": 6072.65,
    "prompt_tokens_mean_ci95_high": 6074.500016842028,
    "prompt_tokens_mean_ci95_low": 6070.799983157971,
    "prompt_tokens_min": 6066.0,
    "prompt_tokens_p25": 6069.0,
    "prompt_tokens_p50": 6072.0,
    "prompt_tokens_p75": 6076.0,
    "prompt_tokens_p90": 6078.0,
    "prompt_tokens_p95": 6078.0,
    "prompt_tokens_p99": 6079.0,
    "prompt_tokens_samples": 20,
    "prompt_tokens_stddev": 4.221187161525958,
    "reasoning_chunks": 3554,
    "request_canary_passed": 0,
    "request_canary_samples": 0,
    "throughput_performance_eligible": true,
    "throughput_tok_s": 83.36275273422383,
    "time_to_first_output_max_ms": 10282.233226927929,
    "time_to_first_output_mean_ci95_high_ms": 5275.154968903333,
    "time_to_first_output_mean_ci95_low_ms": 3719.9190785010087,
    "time_to_first_output_mean_ms": 4497.537023702171,
    "time_to_first_output_min_ms": 2583.0574999563396,
    "time_to_first_output_p25_ms": 3211.002313066274,
    "time_to_first_output_p50_ms": 4276.235112105496,
    "time_to_first_output_p75_ms": 4922.690781997517,
    "time_to_first_output_p90_ms": 5984.028853941709,
    "time_to_first_output_p95_ms": 7046.535634086467,
    "time_to_first_output_p99_ms": 10282.233226927929,
    "time_to_first_output_samples": 20,
    "time_to_first_output_stddev_ms": 1774.2924346872492,
    "tpot_max_ms": 36.80133010284641,
    "tpot_mean_ci95_high_ms": 29.33627144312785,
    "tpot_mean_ci95_low_ms": 25.783812283492708,
    "tpot_mean_ms": 27.56004186331028,
    "tpot_min_ms": 17.83850694918567,
    "tpot_p25_ms": 26.005101269038278,
    "tpot_p50_ms": 26.517781335554133,
    "tpot_p75_ms": 29.562731491212077,
    "tpot_p90_ms": 31.982172222342342,
    "tpot_p95_ms": 33.26490181946537,
    "tpot_p99_ms": 36.80133010284641,
    "tpot_samples": 20,
    "tpot_stddev_ms": 4.052826616446913,
    "ttft_max_ms": 16110.973021015525,
    "ttft_mean_ci95_high_ms": 10742.910851506012,
    "ttft_mean_ci95_low_ms": 8182.390269680918,
    "ttft_mean_ms": 9462.650560593465,
    "ttft_min_ms": 4629.774966975674,
    "ttft_p25_ms": 7413.717303890735,
    "ttft_p50_ms": 8764.1307329759,
    "ttft_p75_ms": 11114.820459973998,
    "ttft_p90_ms": 11694.183403044008,
    "ttft_p95_ms": 15691.028936998919,
    "ttft_p99_ms": 16110.973021015525,
    "ttft_samples": 20,
    "ttft_stddev_ms": 2921.172489157257
  },
  "model": "llm.primary",
  "performance_eligible": true,
  "prompt_cache_mode": "unique",
  "request_canaries": false,
  "request_timings": [
    {
      "content_chunks": 51,
      "controlled_output": null,
      "decode_tok_s": 33.826373598030465,
      "e2e_ms": 16667.783229029737,
      "effective_prefill_tok_s": 590.0469154999577,
      "generation_ms": 6385.550002101809,
      "mean_inter_token_latency_ms": 29.562731491212077,
      "mean_time_per_output_token_ms": 29.562731491212077,
      "output_token_source": "usage",
      "output_tokens": 217,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6067,
      "reasoning_chunks": 161,
      "request_canary": null,
      "request_index": 0,
      "time_to_first_output_ms": 10282.233226927929,
      "tpot_ms": 29.562731491212077,
      "ttft_ms": 15691.028936998919,
      "visible_content_truncated": false,
      "visible_generation_ms": 976.7542920308188
    },
    {
      "content_chunks": 46,
      "controlled_output": null,
      "decode_tok_s": 34.9314874887833,
      "e2e_ms": 8971.13967500627,
      "effective_prefill_tok_s": 1418.537531490703,
      "generation_ms": 4694.904562900774,
      "mean_inter_token_latency_ms": 28.62746684695594,
      "mean_time_per_output_token_ms": 28.62746684695594,
      "output_token_source": "usage",
      "output_tokens": 165,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6066,
      "reasoning_chunks": 110,
      "request_canary": null,
      "request_index": 1,
      "time_to_first_output_ms": 4276.235112105496,
      "tpot_ms": 28.62746684695594,
      "ttft_ms": 8085.840187035501,
      "visible_content_truncated": false,
      "visible_generation_ms": 885.2994879707694
    },
    {
      "content_chunks": 52,
      "controlled_output": null,
      "decode_tok_s": 37.88828223140551,
      "e2e_ms": 12107.699107029475,
      "effective_prefill_tok_s": 1205.139475601241,
      "generation_ms": 7073.427039082162,
      "mean_inter_token_latency_ms": 26.393384474187172,
      "mean_time_per_output_token_ms": 26.393384474187172,
      "output_token_source": "usage",
      "output_tokens": 269,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6067,
      "reasoning_chunks": 208,
      "request_canary": null,
      "request_index": 2,
      "time_to_first_output_ms": 5034.272067947313,
      "tpot_ms": 26.393384474187172,
      "ttft_ms": 11114.820459973998,
      "visible_content_truncated": false,
      "visible_generation_ms": 992.8786470554769
    },
    {
      "content_chunks": 76,
      "controlled_output": null,
      "decode_tok_s": 38.4539936858697,
      "e2e_ms": 14305.661260033958,
      "effective_prefill_tok_s": 1014.0325436437324,
      "generation_ms": 8321.632406092249,
      "mean_inter_token_latency_ms": 26.005101269038278,
      "mean_time_per_output_token_ms": 26.005101269038278,
      "output_token_source": "usage",
      "output_tokens": 321,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6068,
      "reasoning_chunks": 234,
      "request_canary": null,
      "request_index": 3,
      "time_to_first_output_ms": 5984.028853941709,
      "tpot_ms": 26.005101269038278,
      "ttft_ms": 11692.825609003194,
      "visible_content_truncated": false,
      "visible_generation_ms": 2612.835651030764
    },
    {
      "content_chunks": 49,
      "controlled_output": null,
      "decode_tok_s": 34.49563731054698,
      "e2e_ms": 11045.437581953593,
      "effective_prefill_tok_s": 1368.1533900152147,
      "generation_ms": 6609.531458932906,
      "mean_inter_token_latency_ms": 28.989173065495205,
      "mean_time_per_output_token_ms": 28.989173065495205,
      "output_token_source": "usage",
      "output_tokens": 229,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6069,
      "reasoning_chunks": 172,
      "request_canary": null,
      "request_index": 4,
      "time_to_first_output_ms": 4435.906123020686,
      "tpot_ms": 28.989173065495205,
      "ttft_ms": 10109.751585056074,
      "visible_content_truncated": false,
      "visible_generation_ms": 935.6859968975186
    },
    {
      "content_chunks": 64,
      "controlled_output": null,
      "decode_tok_s": 31.267419643916888,
      "e2e_ms": 11565.550553030334,
      "effective_prefill_tok_s": 1731.0112098326158,
      "generation_ms": 8059.50740003027,
      "mean_inter_token_latency_ms": 31.982172222342342,
      "mean_time_per_output_token_ms": 31.982172222342342,
      "output_token_source": "usage",
      "output_tokens": 253,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6069,
      "reasoning_chunks": 178,
      "request_canary": null,
      "request_index": 5,
      "time_to_first_output_ms": 3506.043153000064,
      "tpot_ms": 31.982172222342342,
      "ttft_ms": 9351.663887966424,
      "visible_content_truncated": false,
      "visible_generation_ms": 2213.88666506391
    },
    {
      "content_chunks": 42,
      "controlled_output": null,
      "decode_tok_s": 39.65967105504367,
      "e2e_ms": 8218.740848940797,
      "effective_prefill_tok_s": 1661.0528670506028,
      "generation_ms": 4563.830086961389,
      "mean_inter_token_latency_ms": 25.214530867189993,
      "mean_time_per_output_token_ms": 25.214530867189993,
      "output_token_source": "usage",
      "output_tokens": 182,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6071,
      "reasoning_chunks": 135,
      "request_canary": null,
      "request_index": 6,
      "time_to_first_output_ms": 3654.9107619794086,
      "tpot_ms": 25.214530867189993,
      "ttft_ms": 7413.717303890735,
      "visible_content_truncated": false,
      "visible_generation_ms": 805.0235450500622
    },
    {
      "content_chunks": 60,
      "controlled_output": null,
      "decode_tok_s": 37.710545514576445,
      "e2e_ms": 12727.573451003991,
      "effective_prefill_tok_s": 1301.0676207084448,
      "generation_ms": 8061.405526008457,
      "mean_inter_token_latency_ms": 26.517781335554133,
      "mean_time_per_output_token_ms": 26.517781335554133,
      "output_token_source": "usage",
      "output_tokens": 305,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6071,
      "reasoning_chunks": 234,
      "request_canary": null,
      "request_index": 7,
      "time_to_first_output_ms": 4666.167924995534,
      "tpot_ms": 26.517781335554133,
      "ttft_ms": 11571.897072950378,
      "visible_content_truncated": false,
      "visible_generation_ms": 1155.676378053613
    },
    {
      "content_chunks": 59,
      "controlled_output": null,
      "decode_tok_s": 35.95215193716782,
      "e2e_ms": 11059.526075026952,
      "effective_prefill_tok_s": 1586.0733080161942,
      "generation_ms": 7231.834145961329,
      "mean_inter_token_latency_ms": 27.81474671523588,
      "mean_time_per_output_token_ms": 27.81474671523588,
      "output_token_source": "usage",
      "output_tokens": 261,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6071,
      "reasoning_chunks": 193,
      "request_canary": null,
      "request_index": 8,
      "time_to_first_output_ms": 3827.6919290656224,
      "tpot_ms": 27.81474671523588,
      "ttft_ms": 8764.1307329759,
      "visible_content_truncated": false,
      "visible_generation_ms": 2295.3953420510516
    },
    {
      "content_chunks": 56,
      "controlled_output": null,
      "decode_tok_s": 36.33154852272061,
      "e2e_ms": 9932.042359025218,
      "effective_prefill_tok_s": 2350.7026073181232,
      "generation_ms": 7348.984859068878,
      "mean_inter_token_latency_ms": 27.52428786168119,
      "mean_time_per_output_token_ms": 27.52428786168119,
      "output_token_source": "usage",
      "output_tokens": 268,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6072,
      "reasoning_chunks": 207,
      "request_canary": null,
      "request_index": 9,
      "time_to_first_output_ms": 2583.0574999563396,
      "tpot_ms": 27.52428786168119,
      "ttft_ms": 6534.434804925695,
      "visible_content_truncated": false,
      "visible_generation_ms": 3397.6075540995225
    },
    {
      "content_chunks": 37,
      "controlled_output": null,
      "decode_tok_s": 30.061715060130965,
      "e2e_ms": 12402.184827020392,
      "effective_prefill_tok_s": 861.7000346422276,
      "generation_ms": 5355.6491929339245,
      "mean_inter_token_latency_ms": 33.26490181946537,
      "mean_time_per_output_token_ms": 33.26490181946537,
      "output_token_source": "usage",
      "output_tokens": 162,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6072,
      "reasoning_chunks": 119,
      "request_canary": null,
      "request_index": 10,
      "time_to_first_output_ms": 7046.535634086467,
      "tpot_ms": 33.26490181946537,
      "ttft_ms": 11694.183403044008,
      "visible_content_truncated": false,
      "visible_generation_ms": 708.0014239763841
    },
    {
      "content_chunks": 68,
      "controlled_output": null,
      "decode_tok_s": 27.172930902371235,
      "e2e_ms": 10173.091002972797,
      "effective_prefill_tok_s": 2026.4735373495985,
      "generation_ms": 7176.25937005505,
      "mean_inter_token_latency_ms": 36.80133010284641,
      "mean_time_per_output_token_ms": 36.80133010284641,
      "output_token_source": "usage",
      "output_tokens": 196,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6073,
      "reasoning_chunks": 123,
      "request_canary": null,
      "request_index": 11,
      "time_to_first_output_ms": 2996.831632917747,
      "tpot_ms": 36.80133010284641,
      "ttft_ms": 6538.815682986751,
      "visible_content_truncated": false,
      "visible_generation_ms": 3634.2753199860454
    },
    {
      "content_chunks": 58,
      "controlled_output": null,
      "decode_tok_s": 38.368064265470956,
      "e2e_ms": 6906.332648941316,
      "effective_prefill_tok_s": 2263.352805800933,
      "generation_ms": 4222.261484945193,
      "mean_inter_token_latency_ms": 26.063342499661683,
      "mean_time_per_output_token_ms": 26.063342499661683,
      "output_token_source": "usage",
      "output_tokens": 163,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6075,
      "reasoning_chunks": 101,
      "request_canary": null,
      "request_index": 12,
      "time_to_first_output_ms": 2684.071163996123,
      "tpot_ms": 26.063342499661683,
      "ttft_ms": 4629.774966975674,
      "visible_content_truncated": false,
      "visible_generation_ms": 2276.5576819656417
    },
    {
      "content_chunks": 46,
      "controlled_output": null,
      "decode_tok_s": 33.295157566507406,
      "e2e_ms": 11139.811082044616,
      "effective_prefill_tok_s": 1234.2843109748399,
      "generation_ms": 6217.1203000471,
      "mean_inter_token_latency_ms": 30.034397584768598,
      "mean_time_per_output_token_ms": 30.034397584768598,
      "output_token_source": "usage",
      "output_tokens": 208,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6076,
      "reasoning_chunks": 154,
      "request_canary": null,
      "request_index": 13,
      "time_to_first_output_ms": 4922.690781997517,
      "tpot_ms": 30.034397584768598,
      "ttft_ms": 10267.419226001948,
      "visible_content_truncated": false,
      "visible_generation_ms": 872.3918560426682
    },
    {
      "content_chunks": 65,
      "controlled_output": null,
      "decode_tok_s": 33.00826601666521,
      "e2e_ms": 18510.20039105788,
      "effective_prefill_tok_s": 1892.244043324236,
      "generation_ms": 15299.198077991605,
      "mean_inter_token_latency_ms": 30.295441738597237,
      "mean_time_per_output_token_ms": 30.295441738597237,
      "output_token_source": "usage",
      "output_tokens": 506,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6076,
      "reasoning_chunks": 432,
      "request_canary": null,
      "request_index": 14,
      "time_to_first_output_ms": 3211.002313066274,
      "tpot_ms": 30.295441738597237,
      "ttft_ms": 16110.973021015525,
      "visible_content_truncated": false,
      "visible_generation_ms": 2399.227370042354
    },
    {
      "content_chunks": 48,
      "controlled_output": null,
      "decode_tok_s": 42.96463718119075,
      "e2e_ms": 8870.98974303808,
      "effective_prefill_tok_s": 2103.258472760191,
      "generation_ms": 5981.6634530434385,
      "mean_inter_token_latency_ms": 23.27495507020793,
      "mean_time_per_output_token_ms": 23.27495507020793,
      "output_token_source": "usage",
      "output_tokens": 258,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6077,
      "reasoning_chunks": 196,
      "request_canary": null,
      "request_index": 15,
      "time_to_first_output_ms": 2889.326289994642,
      "tpot_ms": 23.27495507020793,
      "ttft_ms": 7945.885607041419,
      "visible_content_truncated": false,
      "visible_generation_ms": 925.1041359966621
    },
    {
      "content_chunks": 61,
      "controlled_output": null,
      "decode_tok_s": 38.4099405524398,
      "e2e_ms": 9599.013781989925,
      "effective_prefill_tok_s": 1134.9460415112328,
      "generation_ms": 4243.693107971922,
      "mean_inter_token_latency_ms": 26.034927042772527,
      "mean_time_per_output_token_ms": 26.034927042772527,
      "output_token_source": "usage",
      "output_tokens": 164,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6078,
      "reasoning_chunks": 100,
      "request_canary": null,
      "request_index": 16,
      "time_to_first_output_ms": 5355.320674018003,
      "tpot_ms": 26.034927042772527,
      "ttft_ms": 7260.346419992857,
      "visible_content_truncated": false,
      "visible_generation_ms": 2338.667361997068
    },
    {
      "content_chunks": 46,
      "controlled_output": null,
      "decode_tok_s": 37.957732307055,
      "e2e_ms": 8638.504892005585,
      "effective_prefill_tok_s": 1326.8984122367126,
      "generation_ms": 4057.1443719090894,
      "mean_inter_token_latency_ms": 26.345093324084996,
      "mean_time_per_output_token_ms": 26.345093324084996,
      "output_token_source": "usage",
      "output_tokens": 155,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6079,
      "reasoning_chunks": 106,
      "request_canary": null,
      "request_index": 17,
      "time_to_first_output_ms": 4581.360520096496,
      "tpot_ms": 26.345093324084996,
      "ttft_ms": 7763.834936078638,
      "visible_content_truncated": false,
      "visible_generation_ms": 874.6699559269473
    },
    {
      "content_chunks": 49,
      "controlled_output": null,
      "decode_tok_s": 44.2153793306485,
      "e2e_ms": 10100.661977892742,
      "effective_prefill_tok_s": 1673.3086982612651,
      "generation_ms": 6468.337585916743,
      "mean_inter_token_latency_ms": 22.616564985722878,
      "mean_time_per_output_token_ms": 22.616564985722878,
      "output_token_source": "usage",
      "output_tokens": 287,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6078,
      "reasoning_chunks": 227,
      "request_canary": null,
      "request_index": 18,
      "time_to_first_output_ms": 3632.324391975999,
      "tpot_ms": 22.616564985722878,
      "ttft_ms": 9186.415037955157,
      "visible_content_truncated": false,
      "visible_generation_ms": 914.2469399375841
    },
    {
      "content_chunks": 45,
      "controlled_output": null,
      "decode_tok_s": 56.05850326199245,
      "e2e_ms": 8251.686426927336,
      "effective_prefill_tok_s": 1387.43986018916,
      "generation_ms": 3870.956007973291,
      "mean_inter_token_latency_ms": 17.83850694918567,
      "mean_time_per_output_token_ms": 17.83850694918567,
      "output_token_source": "usage",
      "output_tokens": 218,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6078,
      "reasoning_chunks": 164,
      "request_canary": null,
      "request_index": 19,
      "time_to_first_output_ms": 4380.730418954045,
      "tpot_ms": 17.83850694918567,
      "ttft_ms": 7525.252330000512,
      "visible_content_truncated": false,
      "visible_generation_ms": 726.4340969268233
    }
  ],
  "requests": 20,
  "response_words": 0,
  "responses_received": 20,
  "run_id": "benchmark-20261017T001932Z",
  "sampling": {
    "temperature": {
      "effective_request": 0.6,
      "requested": 0.6,
      "sent": true
    },
    "top_p": {
      "effective_request": 0.95,
      "requested": 0.95,
      "sent": true
    }
  },
  "schema": "anvil-serving.benchmark/v1",
  "serve_flags": {
    "chat_template_kwargs": null,
    "force_nonempty_content": false,
    "no_thinking": false,
    "reasoning_effort": "max",
    "shared_prefix_burst": false,
    "thinking_mode": "default"
  },
  "shared_prefix_tokens": 0,
  "source_recipe": null,
  "started_at": "2026-10-17T00:19:32Z",
  "started_at_unix_ns": 1792196372731436897,
  "tier": null,
  "timestamp_precision": "nanosecond-unix-fields; ISO labels are second precision",
  "timing_methodology": {
    "artifact_timestamps": "time.time_ns Unix timestamps; independent artifacts may align them only when the operator supplied the same explicit clock_domain_id",
    "clock": "client time.perf_counter",
    "controlled_output": "--response-words adherence is recorded; observe preserves historical behavior while strict rejects non-exact or truncated completions; visible response capture is capped at 8192 characters",
    "decode": "usage completion tokens after the first token divided by client-observed generation time; completion tokens may include reasoning tokens",
    "distributions": "request-level mean, sample standard deviation, normal-approximation 95% confidence interval for the 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": "legacy key for a TPOT proxy: client-observed generation time divided by completion tokens after the first token; not raw per-token ITL",
    "p99": "numeric nearest-rank p99 is always retained for compatibility; it is descriptive only below 100 eligible samples",
    "request_canaries": "when enabled, each response must begin with its own synthetic marker and contain no other ANVIL_REQ marker; bounded/truncated captures fail closed and are excluded from performance metrics",
    "time_to_first_output": "request start through the first non-empty streamed reasoning or content delta",
    "tpot": "client-observed mean time per output token after the first output; identical to the legacy mean_inter_token_latency key",
    "ttft": "request start through the first non-empty streamed content delta"
  },
  "unique_prefix_namespace": "capacity-v1:seed-request-index",
  "wall_clock_ms": 57423.727540066466
}