{
  "base_url": "redacted://routed-capability",
  "clock_domain_id": null,
  "completed": 20,
  "concurrency": 1,
  "configuration_fingerprint": null,
  "context_distribution": {
    "32768": 20
  },
  "context_seed": 941121,
  "context_tokens": 32768,
  "controlled_output_policy": "observe",
  "engine": "vllm-exl3-v84",
  "failed": 0,
  "failures": [],
  "finished_at": "2026-10-17T00:26:20Z",
  "finished_at_unix_ns": 1792196780120505915,
  "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": 12833.0,
    "completion_tokens_mean": 831.4,
    "completion_tokens_mean_ci95_high": 2069.9631345713296,
    "completion_tokens_mean_ci95_low": -407.1631345713298,
    "completion_tokens_min": 102.0,
    "completion_tokens_p25": 114.0,
    "completion_tokens_p50": 206.0,
    "completion_tokens_p75": 264.0,
    "completion_tokens_p90": 331.0,
    "completion_tokens_p95": 352.0,
    "completion_tokens_p99": 12833.0,
    "completion_tokens_samples": 20,
    "completion_tokens_stddev": 2826.0320033946055,
    "content_chunks": 1117,
    "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": 68.31229968764521,
    "decode_tok_s_mean": 68.01532297485782,
    "decode_tok_s_mean_ci95_high": 68.09331935926201,
    "decode_tok_s_mean_ci95_low": 67.93732659045364,
    "decode_tok_s_min": 67.72558242043613,
    "decode_tok_s_p25": 67.89440583607775,
    "decode_tok_s_p50": 68.0275710155947,
    "decode_tok_s_p75": 68.14371939422236,
    "decode_tok_s_p90": 68.20747422852659,
    "decode_tok_s_p95": 68.21470763176603,
    "decode_tok_s_p99": 68.31229968764521,
    "decode_tok_s_samples": 20,
    "decode_tok_s_stddev": 0.17796450768058458,
    "e2e_max_ms": 193448.50961398333,
    "e2e_mean_ci95_high_ms": 35511.812557311045,
    "e2e_mean_ci95_low_ms": -837.7270535060852,
    "e2e_mean_ms": 17337.04275190248,
    "e2e_min_ms": 6638.9124220004305,
    "e2e_p25_ms": 6790.759621071629,
    "e2e_p50_ms": 8176.866049994715,
    "e2e_p75_ms": 9048.071398050524,
    "e2e_p90_ms": 10013.32173193805,
    "e2e_p95_ms": 10310.306339059025,
    "e2e_p99_ms": 193448.50961398333,
    "e2e_samples": 20,
    "e2e_stddev_ms": 41469.40893908589,
    "effective_prefill_tok_s_max": 4955.843154162108,
    "effective_prefill_tok_s_mean": 4912.935535386846,
    "effective_prefill_tok_s_mean_ci95_high": 4923.740046332339,
    "effective_prefill_tok_s_mean_ci95_low": 4902.131024441353,
    "effective_prefill_tok_s_min": 4858.887034428073,
    "effective_prefill_tok_s_p25": 4895.492036519277,
    "effective_prefill_tok_s_p50": 4912.236939395865,
    "effective_prefill_tok_s_p75": 4926.0838968796215,
    "effective_prefill_tok_s_p90": 4947.285020828527,
    "effective_prefill_tok_s_p95": 4949.990663069202,
    "effective_prefill_tok_s_p99": 4955.843154162108,
    "effective_prefill_tok_s_samples": 20,
    "effective_prefill_tok_s_stddev": 24.652674426289877,
    "generation_max_ms": 188307.88976699114,
    "generation_mean_ci95_high_ms": 30365.830620065757,
    "generation_mean_ci95_low_ms": -5984.738531745465,
    "generation_mean_ms": 12190.546044160146,
    "generation_min_ms": 1490.9411480184644,
    "generation_p25_ms": 1663.9541510958225,
    "generation_p50_ms": 3006.986398017034,
    "generation_p75_ms": 3865.134426043369,
    "generation_p90_ms": 4851.746864034794,
    "generation_p95_ms": 5151.75720595289,
    "generation_p99_ms": 188307.88976699114,
    "generation_samples": 20,
    "generation_stddev_ms": 41470.583491967736,
    "mean_inter_token_latency_ms_max": 14.76546918108527,
    "mean_inter_token_latency_ms_mean": 14.702665018510407,
    "mean_inter_token_latency_ms_mean_ci95_high": 14.719540443272148,
    "mean_inter_token_latency_ms_mean_ci95_low": 14.685789593748666,
    "mean_inter_token_latency_ms_min": 14.638652256950111,
    "mean_inter_token_latency_ms_p25": 14.668825510848976,
    "mean_inter_token_latency_ms_p50": 14.696328616134483,
    "mean_inter_token_latency_ms_p75": 14.725257974299314,
    "mean_inter_token_latency_ms_p90": 14.761793544737271,
    "mean_inter_token_latency_ms_p95": 14.765337599568408,
    "mean_inter_token_latency_ms_p99": 14.76546918108527,
    "mean_inter_token_latency_ms_samples": 20,
    "mean_inter_token_latency_ms_stddev": 0.03850469073105682,
    "mean_time_per_output_token_max_ms": 14.76546918108527,
    "mean_time_per_output_token_mean_ci95_high_ms": 14.719540443272148,
    "mean_time_per_output_token_mean_ci95_low_ms": 14.685789593748666,
    "mean_time_per_output_token_mean_ms": 14.702665018510407,
    "mean_time_per_output_token_min_ms": 14.638652256950111,
    "mean_time_per_output_token_p25_ms": 14.668825510848976,
    "mean_time_per_output_token_p50_ms": 14.696328616134483,
    "mean_time_per_output_token_p75_ms": 14.725257974299314,
    "mean_time_per_output_token_p90_ms": 14.761793544737271,
    "mean_time_per_output_token_p95_ms": 14.765337599568408,
    "mean_time_per_output_token_p99_ms": 14.76546918108527,
    "mean_time_per_output_token_samples": 20,
    "mean_time_per_output_token_stddev_ms": 0.03850469073105682,
    "output_token_sources": {
      "usage": 20
    },
    "output_tokens": 16628,
    "prefix_cache_hit_avg": null,
    "prompt_token_samples": 20,
    "prompt_tokens": 505676,
    "prompt_tokens_max": 25287.0,
    "prompt_tokens_mean": 25283.8,
    "prompt_tokens_mean_ci95_high": 25284.77067139,
    "prompt_tokens_mean_ci95_low": 25282.829328609998,
    "prompt_tokens_min": 25280.0,
    "prompt_tokens_p25": 25281.0,
    "prompt_tokens_p50": 25284.0,
    "prompt_tokens_p75": 25285.0,
    "prompt_tokens_p90": 25286.0,
    "prompt_tokens_p95": 25286.0,
    "prompt_tokens_p99": 25287.0,
    "prompt_tokens_samples": 20,
    "prompt_tokens_stddev": 2.2147828692435603,
    "reasoning_chunks": 14438,
    "request_canary_passed": 0,
    "request_canary_samples": 0,
    "throughput_performance_eligible": true,
    "throughput_tok_s": 47.94982955986604,
    "time_to_first_output_max_ms": 5203.866609954275,
    "time_to_first_output_mean_ci95_high_ms": 5157.84909307634,
    "time_to_first_output_mean_ci95_low_ms": 5135.1443224083305,
    "time_to_first_output_mean_ms": 5146.496707742335,
    "time_to_first_output_min_ms": 5101.049249060452,
    "time_to_first_output_p25_ms": 5128.708604956046,
    "time_to_first_output_p50_ms": 5140.619846992195,
    "time_to_first_output_p75_ms": 5161.574867903255,
    "time_to_first_output_p90_ms": 5169.879651977681,
    "time_to_first_output_p95_ms": 5185.730154043995,
    "time_to_first_output_p99_ms": 5203.866609954275,
    "time_to_first_output_samples": 20,
    "time_to_first_output_stddev_ms": 25.902760524087796,
    "tpot_max_ms": 14.76546918108527,
    "tpot_mean_ci95_high_ms": 14.719540443272148,
    "tpot_mean_ci95_low_ms": 14.685789593748666,
    "tpot_mean_ms": 14.702665018510407,
    "tpot_min_ms": 14.638652256950111,
    "tpot_p25_ms": 14.668825510848976,
    "tpot_p50_ms": 14.696328616134483,
    "tpot_p75_ms": 14.725257974299314,
    "tpot_p90_ms": 14.761793544737271,
    "tpot_p95_ms": 14.765337599568408,
    "tpot_p99_ms": 14.76546918108527,
    "tpot_samples": 20,
    "tpot_stddev_ms": 0.03850469073105682,
    "ttft_max_ms": 192218.23533892166,
    "ttft_mean_ci95_high_ms": 34634.20871012242,
    "ttft_mean_ci95_low_ms": -1632.241272347539,
    "ttft_mean_ms": 16500.983718887437,
    "ttft_min_ms": 5975.042192032561,
    "ttft_p25_ms": 6068.618518067524,
    "ttft_p50_ms": 7358.747574035078,
    "ttft_p75_ms": 8070.980365970172,
    "ttft_p90_ms": 9108.698628959246,
    "ttft_p95_ms": 9290.20308109466,
    "ttft_p99_ms": 192218.23533892166,
    "ttft_samples": 20,
    "ttft_stddev_ms": 41374.61605275453
  },
  "model": "llm.primary",
  "performance_eligible": true,
  "prompt_cache_mode": "unique",
  "request_canaries": false,
  "request_timings": [
    {
      "content_chunks": 42,
      "controlled_output": null,
      "decode_tok_s": 68.20747422852659,
      "e2e_ms": 6684.453431982547,
      "effective_prefill_tok_s": 4955.843154162108,
      "generation_ms": 1583.404182922095,
      "mean_inter_token_latency_ms": 14.66114984187125,
      "mean_time_per_output_token_ms": 14.66114984187125,
      "output_token_source": "usage",
      "output_tokens": 109,
      "planned_context_tokens": 32768,
      "prompt_tokens": 25280,
      "reasoning_chunks": 64,
      "request_canary": null,
      "request_index": 0,
      "time_to_first_output_ms": 5101.049249060452,
      "tpot_ms": 14.66114984187125,
      "ttft_ms": 6048.9202729659155,
      "visible_content_truncated": false,
      "visible_generation_ms": 635.5331590166315
    },
    {
      "content_chunks": 55,
      "controlled_output": null,
      "decode_tok_s": 68.17456844340495,
      "e2e_ms": 8176.866049994715,
      "effective_prefill_tok_s": 4890.055804360752,
      "generation_ms": 3006.986398017034,
      "mean_inter_token_latency_ms": 14.66822633179041,
      "mean_time_per_output_token_ms": 14.66822633179041,
      "output_token_source": "usage",
      "output_tokens": 206,
      "planned_context_tokens": 32768,
      "prompt_tokens": 25281,
      "reasoning_chunks": 145,
      "request_canary": null,
      "request_index": 1,
      "time_to_first_output_ms": 5169.879651977681,
      "tpot_ms": 14.66822633179041,
      "ttft_ms": 7358.747574035078,
      "visible_content_truncated": false,
      "visible_generation_ms": 818.1184759596363
    },
    {
      "content_chunks": 44,
      "controlled_output": null,
      "decode_tok_s": 67.7424458599416,
      "e2e_ms": 6638.9124220004305,
      "effective_prefill_tok_s": 4910.866563644419,
      "generation_ms": 1490.9411480184644,
      "mean_inter_token_latency_ms": 14.761793544737271,
      "mean_time_per_output_token_ms": 14.761793544737271,
      "output_token_source": "usage",
      "output_tokens": 102,
      "planned_context_tokens": 32768,
      "prompt_tokens": 25281,
      "reasoning_chunks": 56,
      "request_canary": null,
      "request_index": 2,
      "time_to_first_output_ms": 5147.971273981966,
      "tpot_ms": 14.761793544737271,
      "ttft_ms": 5975.042192032561,
      "visible_content_truncated": false,
      "visible_generation_ms": 663.8702299678698
    },
    {
      "content_chunks": 44,
      "controlled_output": null,
      "decode_tok_s": 67.72558242043613,
      "e2e_ms": 6736.104418057948,
      "effective_prefill_tok_s": 4874.916212191617,
      "generation_ms": 1550.3742640139535,
      "mean_inter_token_latency_ms": 14.76546918108527,
      "mean_time_per_output_token_ms": 14.76546918108527,
      "output_token_source": "usage",
      "output_tokens": 106,
      "planned_context_tokens": 32768,
      "prompt_tokens": 25280,
      "reasoning_chunks": 59,
      "request_canary": null,
      "request_index": 3,
      "time_to_first_output_ms": 5185.730154043995,
      "tpot_ms": 14.76546918108527,
      "ttft_ms": 6068.618518067524,
      "visible_content_truncated": false,
      "visible_generation_ms": 667.4858999904245
    },
    {
      "content_chunks": 49,
      "controlled_output": null,
      "decode_tok_s": 68.0275710155947,
      "e2e_ms": 8498.117391951382,
      "effective_prefill_tok_s": 4912.236939395865,
      "generation_ms": 3351.5822569606826,
      "mean_inter_token_latency_ms": 14.699922179652116,
      "mean_time_per_output_token_ms": 14.699922179652116,
      "output_token_source": "usage",
      "output_tokens": 229,
      "planned_context_tokens": 32768,
      "prompt_tokens": 25281,
      "reasoning_chunks": 172,
      "request_canary": null,
      "request_index": 4,
      "time_to_first_output_ms": 5146.5351349907,
      "tpot_ms": 14.699922179652116,
      "ttft_ms": 7765.740755945444,
      "visible_content_truncated": false,
      "visible_generation_ms": 732.376636005938
    },
    {
      "content_chunks": 57,
      "controlled_output": null,
      "decode_tok_s": 68.31229968764521,
      "e2e_ms": 7758.9777410030365,
      "effective_prefill_tok_s": 4919.960648082579,
      "generation_ms": 2620.31875399407,
      "mean_inter_token_latency_ms": 14.638652256950111,
      "mean_time_per_output_token_ms": 14.638652256950111,
      "output_token_source": "usage",
      "output_tokens": 180,
      "planned_context_tokens": 32768,
      "prompt_tokens": 25282,
      "reasoning_chunks": 118,
      "request_canary": null,
      "request_index": 5,
      "time_to_first_output_ms": 5138.6589870089665,
      "tpot_ms": 14.638652256950111,
      "ttft_ms": 6915.135326911695,
      "visible_content_truncated": false,
      "visible_generation_ms": 843.8424140913412
    },
    {
      "content_chunks": 44,
      "controlled_output": null,
      "decode_tok_s": 67.72618595793097,
      "e2e_ms": 6731.8735379958525,
      "effective_prefill_tok_s": 4949.990663069202,
      "generation_ms": 1624.1871359525248,
      "mean_inter_token_latency_ms": 14.765337599568408,
      "mean_time_per_output_token_ms": 14.765337599568408,
      "output_token_source": "usage",
      "output_tokens": 111,
      "planned_context_tokens": 32768,
      "prompt_tokens": 25283,
      "reasoning_chunks": 63,
      "request_canary": null,
      "request_index": 6,
      "time_to_first_output_ms": 5107.686402043328,
      "tpot_ms": 14.765337599568408,
      "ttft_ms": 6065.957247978076,
      "visible_content_truncated": false,
      "visible_generation_ms": 665.9162900177762
    },
    {
      "content_chunks": 50,
      "controlled_output": null,
      "decode_tok_s": 67.91052501391466,
      "e2e_ms": 6790.759621071629,
      "effective_prefill_tok_s": 4931.726032530607,
      "generation_ms": 1663.9541510958225,
      "mean_inter_token_latency_ms": 14.725257974299314,
      "mean_time_per_output_token_ms": 14.725257974299314,
      "output_token_source": "usage",
      "output_tokens": 114,
      "planned_context_tokens": 32768,
      "prompt_tokens": 25284,
      "reasoning_chunks": 61,
      "request_canary": null,
      "request_index": 7,
      "time_to_first_output_ms": 5126.805469975807,
      "tpot_ms": 14.725257974299314,
      "ttft_ms": 6040.5686480225995,
      "visible_content_truncated": false,
      "visible_generation_ms": 750.1909730490297
    },
    {
      "content_chunks": 77,
      "controlled_output": null,
      "decode_tok_s": 68.04420519708181,
      "e2e_ms": 8975.816324003972,
      "effective_prefill_tok_s": 4947.285020828527,
      "generation_ms": 3865.134426043369,
      "mean_inter_token_latency_ms": 14.696328616134483,
      "mean_time_per_output_token_ms": 14.696328616134483,
      "output_token_source": "usage",
      "output_tokens": 264,
      "planned_context_tokens": 32768,
      "prompt_tokens": 25284,
      "reasoning_chunks": 182,
      "request_canary": null,
      "request_index": 8,
      "time_to_first_output_ms": 5110.681897960603,
      "tpot_ms": 14.696328616134483,
      "ttft_ms": 7830.2876349771395,
      "visible_content_truncated": false,
      "visible_generation_ms": 1145.5286890268326
    },
    {
      "content_chunks": 83,
      "controlled_output": null,
      "decode_tok_s": 68.14371939422236,
      "e2e_ms": 193448.50961398333,
      "effective_prefill_tok_s": 4918.473015427081,
      "generation_ms": 188307.88976699114,
      "mean_inter_token_latency_ms": 14.674866721243076,
      "mean_time_per_output_token_ms": 14.674866721243076,
      "output_token_source": "usage",
      "output_tokens": 12833,
      "planned_context_tokens": 32768,
      "prompt_tokens": 25284,
      "reasoning_chunks": 11777,
      "request_canary": null,
      "request_index": 9,
      "time_to_first_output_ms": 5140.619846992195,
      "tpot_ms": 14.674866721243076,
      "ttft_ms": 192218.23533892166,
      "visible_content_truncated": false,
      "visible_generation_ms": 1230.2742750616744
    },
    {
      "content_chunks": 47,
      "controlled_output": null,
      "decode_tok_s": 67.95099825778016,
      "e2e_ms": 9528.93829497043,
      "effective_prefill_tok_s": 4930.090973693893,
      "generation_ms": 4400.229690014385,
      "mean_inter_token_latency_ms": 14.716487257573194,
      "mean_time_per_output_token_ms": 14.716487257573194,
      "output_token_source": "usage",
      "output_tokens": 300,
      "planned_context_tokens": 32768,
      "prompt_tokens": 25285,
      "reasoning_chunks": 244,
      "request_canary": null,
      "request_index": 10,
      "time_to_first_output_ms": 5128.708604956046,
      "tpot_ms": 14.716487257573194,
      "ttft_ms": 8815.483703976497,
      "visible_content_truncated": false,
      "visible_generation_ms": 713.4545909939334
    },
    {
      "content_chunks": 69,
      "controlled_output": null,
      "decode_tok_s": 68.13209279241987,
      "e2e_ms": 10310.306339059025,
      "effective_prefill_tok_s": 4901.76585461239,
      "generation_ms": 5151.75720595289,
      "mean_inter_token_latency_ms": 14.67737095713074,
      "mean_time_per_output_token_ms": 14.67737095713074,
      "output_token_source": "usage",
      "output_tokens": 352,
      "planned_context_tokens": 32768,
      "prompt_tokens": 25286,
      "reasoning_chunks": 275,
      "request_canary": null,
      "request_index": 11,
      "time_to_first_output_ms": 5158.549133106135,
      "tpot_ms": 14.67737095713074,
      "ttft_ms": 9290.20308109466,
      "visible_content_truncated": false,
      "visible_generation_ms": 1020.1032579643652
    },
    {
      "content_chunks": 59,
      "controlled_output": null,
      "decode_tok_s": 68.01673897008848,
      "e2e_ms": 10013.32173193805,
      "effective_prefill_tok_s": 4898.892420845913,
      "generation_ms": 4851.746864034794,
      "mean_inter_token_latency_ms": 14.702263224347861,
      "mean_time_per_output_token_ms": 14.702263224347861,
      "output_token_source": "usage",
      "output_tokens": 331,
      "planned_context_tokens": 32768,
      "prompt_tokens": 25286,
      "reasoning_chunks": 260,
      "request_canary": null,
      "request_index": 12,
      "time_to_first_output_ms": 5161.574867903255,
      "tpot_ms": 14.702263224347861,
      "ttft_ms": 9108.698628959246,
      "visible_content_truncated": false,
      "visible_generation_ms": 904.6231029788032
    },
    {
      "content_chunks": 65,
      "controlled_output": null,
      "decode_tok_s": 67.89440583607775,
      "e2e_ms": 9048.071398050524,
      "effective_prefill_tok_s": 4858.887034428073,
      "generation_ms": 3844.204788096249,
      "mean_inter_token_latency_ms": 14.728753977380265,
      "mean_time_per_output_token_ms": 14.728753977380265,
      "output_token_source": "usage",
      "output_tokens": 262,
      "planned_context_tokens": 32768,
      "prompt_tokens": 25285,
      "reasoning_chunks": 190,
      "request_canary": null,
      "request_index": 13,
      "time_to_first_output_ms": 5203.866609954275,
      "tpot_ms": 14.728753977380265,
      "ttft_ms": 8070.980365970172,
      "visible_content_truncated": false,
      "visible_generation_ms": 977.0910320803523
    },
    {
      "content_chunks": 43,
      "controlled_output": null,
      "decode_tok_s": 68.08198930212767,
      "e2e_ms": 8648.559147957712,
      "effective_prefill_tok_s": 4893.117481999065,
      "generation_ms": 3481.096872012131,
      "mean_inter_token_latency_ms": 14.688172455747388,
      "mean_time_per_output_token_ms": 14.688172455747388,
      "output_token_source": "usage",
      "output_tokens": 238,
      "planned_context_tokens": 32768,
      "prompt_tokens": 25285,
      "reasoning_chunks": 190,
      "request_canary": null,
      "request_index": 14,
      "time_to_first_output_ms": 5167.4622759455815,
      "tpot_ms": 14.688172455747388,
      "ttft_ms": 8010.470898938365,
      "visible_content_truncated": false,
      "visible_generation_ms": 638.0882490193471
    },
    {
      "content_chunks": 48,
      "controlled_output": null,
      "decode_tok_s": 67.75207134084428,
      "e2e_ms": 7081.160422996618,
      "effective_prefill_tok_s": 4926.0838968796215,
      "generation_ms": 1948.27991805505,
      "mean_inter_token_latency_ms": 14.759696348901894,
      "mean_time_per_output_token_ms": 14.759696348901894,
      "output_token_source": "usage",
      "output_tokens": 133,
      "planned_context_tokens": 32768,
      "prompt_tokens": 25285,
      "reasoning_chunks": 80,
      "request_canary": null,
      "request_index": 15,
      "time_to_first_output_ms": 5132.880504941568,
      "tpot_ms": 14.759696348901894,
      "ttft_ms": 6361.26727599185,
      "visible_content_truncated": false,
      "visible_generation_ms": 719.8931470047683
    },
    {
      "content_chunks": 54,
      "controlled_output": null,
      "decode_tok_s": 68.10696071814961,
      "e2e_ms": 9081.075637950562,
      "effective_prefill_tok_s": 4899.655023532928,
      "generation_ms": 3920.3041390283033,
      "mean_inter_token_latency_ms": 14.682787037559189,
      "mean_time_per_output_token_ms": 14.682787037559189,
      "output_token_source": "usage",
      "output_tokens": 268,
      "planned_context_tokens": 32768,
      "prompt_tokens": 25286,
      "reasoning_chunks": 209,
      "request_canary": null,
      "request_index": 16,
      "time_to_first_output_ms": 5160.771498922259,
      "tpot_ms": 14.682787037559189,
      "ttft_ms": 8267.549658892676,
      "visible_content_truncated": false,
      "visible_generation_ms": 813.5259790578857
    },
    {
      "content_chunks": 69,
      "controlled_output": null,
      "decode_tok_s": 68.17178370963688,
      "e2e_ms": 8528.24644302018,
      "effective_prefill_tok_s": 4919.502129249592,
      "generation_ms": 3388.498693006113,
      "mean_inter_token_latency_ms": 14.668825510848976,
      "mean_time_per_output_token_ms": 14.668825510848976,
      "output_token_source": "usage",
      "output_tokens": 232,
      "planned_context_tokens": 32768,
      "prompt_tokens": 25285,
      "reasoning_chunks": 158,
      "request_canary": null,
      "request_index": 17,
      "time_to_first_output_ms": 5139.747750014067,
      "tpot_ms": 14.668825510848976,
      "ttft_ms": 7506.534120067954,
      "visible_content_truncated": false,
      "visible_generation_ms": 1021.7123229522258
    },
    {
      "content_chunks": 53,
      "controlled_output": null,
      "decode_tok_s": 68.21470763176603,
      "e2e_ms": 6924.311495036818,
      "effective_prefill_tok_s": 4895.492036519277,
      "generation_ms": 1759.1514230007306,
      "mean_inter_token_latency_ms": 14.659595191672755,
      "mean_time_per_output_token_ms": 14.659595191672755,
      "output_token_source": "usage",
      "output_tokens": 121,
      "planned_context_tokens": 32768,
      "prompt_tokens": 25286,
      "reasoning_chunks": 66,
      "request_canary": null,
      "request_index": 18,
      "time_to_first_output_ms": 5165.1600720360875,
      "tpot_ms": 14.659595191672755,
      "ttft_ms": 6136.696920031682,
      "visible_content_truncated": false,
      "visible_generation_ms": 787.6145750051364
    },
    {
      "content_chunks": 65,
      "controlled_output": null,
      "decode_tok_s": 67.97013371956687,
      "e2e_ms": 7136.473575024866,
      "effective_prefill_tok_s": 4923.869802283404,
      "generation_ms": 2000.8788059931248,
      "mean_inter_token_latency_ms": 14.712344161714153,
      "mean_time_per_output_token_ms": 14.712344161714153,
      "output_token_source": "usage",
      "output_tokens": 137,
      "planned_context_tokens": 32768,
      "prompt_tokens": 25287,
      "reasoning_chunks": 69,
      "request_canary": null,
      "request_index": 19,
      "time_to_first_output_ms": 5135.594769031741,
      "tpot_ms": 14.712344161714153,
      "ttft_ms": 6164.536213967949,
      "visible_content_truncated": false,
      "visible_generation_ms": 971.9373610569164
    }
  ],
  "requests": 20,
  "response_words": 0,
  "responses_received": 20,
  "run_id": "benchmark-20261017T002033Z",
  "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": "2026-10-17T00:20:33Z",
  "started_at_unix_ns": 1792196433341375090,
  "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": 346779.1262790561
}