{
  "base_url": "redacted://routed-capability",
  "clock_domain_id": null,
  "completed": 20,
  "concurrency": 1,
  "configuration_fingerprint": null,
  "context_distribution": {
    "32768": 20
  },
  "context_seed": 941221,
  "context_tokens": 32768,
  "controlled_output_policy": "observe",
  "engine": "vllm-exl3-v84",
  "failed": 0,
  "failures": [],
  "finished_at": "2026-10-17T00:34:58Z",
  "finished_at_unix_ns": 1792197298119254493,
  "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": 14438.0,
    "completion_tokens_mean": 1448.95,
    "completion_tokens_mean_ci95_high": 3045.55146256643,
    "completion_tokens_mean_ci95_low": -147.6514625664297,
    "completion_tokens_min": 54.0,
    "completion_tokens_p25": 165.0,
    "completion_tokens_p50": 237.0,
    "completion_tokens_p75": 381.0,
    "completion_tokens_p90": 1285.0,
    "completion_tokens_p95": 9097.0,
    "completion_tokens_p99": 14438.0,
    "completion_tokens_samples": 20,
    "completion_tokens_stddev": 3642.968778851146,
    "content_chunks": 1218,
    "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": 77.16772980024048,
    "decode_tok_s_mean": 70.52962206680398,
    "decode_tok_s_mean_ci95_high": 72.36825365436884,
    "decode_tok_s_mean_ci95_low": 68.69099047923912,
    "decode_tok_s_min": 66.26370663176357,
    "decode_tok_s_p25": 67.86795138092191,
    "decode_tok_s_p50": 68.11678552574685,
    "decode_tok_s_p75": 75.97270633104887,
    "decode_tok_s_p90": 76.93138051397554,
    "decode_tok_s_p95": 76.96147741408313,
    "decode_tok_s_p99": 77.16772980024048,
    "decode_tok_s_samples": 20,
    "decode_tok_s_stddev": 4.1952094034422425,
    "e2e_max_ms": 223057.7286890475,
    "e2e_mean_ci95_high_ms": 49215.45281340458,
    "e2e_mean_ci95_low_ms": 2561.758046794246,
    "e2e_mean_ms": 25888.605430099415,
    "e2e_min_ms": 5914.998011896387,
    "e2e_p25_ms": 7301.461171009578,
    "e2e_p50_ms": 8612.800211063586,
    "e2e_p75_ms": 10031.399628031068,
    "e2e_p90_ms": 23994.117031921633,
    "e2e_p95_ms": 124872.20033397898,
    "e2e_p99_ms": 223057.7286890475,
    "e2e_samples": 20,
    "e2e_stddev_ms": 53224.91474472801,
    "effective_prefill_tok_s_max": 5003.553450989291,
    "effective_prefill_tok_s_mean": 4924.589599207627,
    "effective_prefill_tok_s_mean_ci95_high": 4941.0711773942185,
    "effective_prefill_tok_s_mean_ci95_low": 4908.108021021035,
    "effective_prefill_tok_s_min": 4875.214005298059,
    "effective_prefill_tok_s_p25": 4898.5388314178235,
    "effective_prefill_tok_s_p50": 4914.681597827631,
    "effective_prefill_tok_s_p75": 4930.410989473405,
    "effective_prefill_tok_s_p90": 4974.490890606123,
    "effective_prefill_tok_s_p95": 4998.633436303636,
    "effective_prefill_tok_s_p99": 5003.553450989291,
    "effective_prefill_tok_s_samples": 20,
    "effective_prefill_tok_s_stddev": 37.60605020581353,
    "generation_max_ms": 217871.90505699255,
    "generation_mean_ci95_high_ms": 44075.17565835506,
    "generation_mean_ci95_low_ms": -2567.321854268721,
    "generation_mean_ms": 20753.926902043168,
    "generation_min_ms": 786.0180599382147,
    "generation_p25_ms": 2215.952191967517,
    "generation_p50_ms": 3477.561639039777,
    "generation_p75_ms": 4944.05333395116,
    "generation_p90_ms": 18826.001756009646,
    "generation_p95_ms": 119727.2078259848,
    "generation_p99_ms": 217871.90505699255,
    "generation_samples": 20,
    "generation_stddev_ms": 53212.140346220214,
    "mean_inter_token_latency_ms_max": 15.09121736212458,
    "mean_inter_token_latency_ms_mean": 14.224024191647079,
    "mean_inter_token_latency_ms_mean_ci95_high": 14.577620888727921,
    "mean_inter_token_latency_ms_mean_ci95_low": 13.870427494566236,
    "mean_inter_token_latency_ms_min": 12.958784748348053,
    "mean_inter_token_latency_ms_p25": 13.068204114805102,
    "mean_inter_token_latency_ms_p50": 14.67719238018617,
    "mean_inter_token_latency_ms_p75": 14.708007530711223,
    "mean_inter_token_latency_ms_p90": 14.739291469861822,
    "mean_inter_token_latency_ms_p95": 14.830529432796505,
    "mean_inter_token_latency_ms_p99": 15.09121736212458,
    "mean_inter_token_latency_ms_samples": 20,
    "mean_inter_token_latency_ms_stddev": 0.8068022972369044,
    "mean_time_per_output_token_max_ms": 15.09121736212458,
    "mean_time_per_output_token_mean_ci95_high_ms": 14.577620888727921,
    "mean_time_per_output_token_mean_ci95_low_ms": 13.870427494566236,
    "mean_time_per_output_token_mean_ms": 14.224024191647079,
    "mean_time_per_output_token_min_ms": 12.958784748348053,
    "mean_time_per_output_token_p25_ms": 13.068204114805102,
    "mean_time_per_output_token_p50_ms": 14.67719238018617,
    "mean_time_per_output_token_p75_ms": 14.708007530711223,
    "mean_time_per_output_token_p90_ms": 14.739291469861822,
    "mean_time_per_output_token_p95_ms": 14.830529432796505,
    "mean_time_per_output_token_p99_ms": 15.09121736212458,
    "mean_time_per_output_token_samples": 20,
    "mean_time_per_output_token_stddev_ms": 0.8068022972369044,
    "output_token_sources": {
      "usage": 20
    },
    "output_tokens": 28979,
    "prefix_cache_hit_avg": null,
    "prompt_token_samples": 20,
    "prompt_tokens": 505696,
    "prompt_tokens_max": 25288.0,
    "prompt_tokens_mean": 25284.8,
    "prompt_tokens_mean_ci95_high": 25285.77067139,
    "prompt_tokens_mean_ci95_low": 25283.829328609998,
    "prompt_tokens_min": 25281.0,
    "prompt_tokens_p25": 25282.0,
    "prompt_tokens_p50": 25285.0,
    "prompt_tokens_p75": 25286.0,
    "prompt_tokens_p90": 25287.0,
    "prompt_tokens_p95": 25287.0,
    "prompt_tokens_p99": 25288.0,
    "prompt_tokens_samples": 20,
    "prompt_tokens_stddev": 2.2147828692435603,
    "reasoning_chunks": 25359,
    "request_canary_passed": 0,
    "request_canary_samples": 0,
    "throughput_performance_eligible": true,
    "throughput_tok_s": 55.964954205256575,
    "time_to_first_output_max_ms": 5185.823632054962,
    "time_to_first_output_mean_ci95_high_ms": 5151.664009554204,
    "time_to_first_output_mean_ci95_low_ms": 5117.69304655829,
    "time_to_first_output_mean_ms": 5134.678528056247,
    "time_to_first_output_min_ms": 5053.808307973668,
    "time_to_first_output_p25_ms": 5098.079220042564,
    "time_to_first_output_p50_ms": 5143.102804082446,
    "time_to_first_output_p75_ms": 5159.616862074472,
    "time_to_first_output_p90_ms": 5175.421694060788,
    "time_to_first_output_p95_ms": 5177.633792976849,
    "time_to_first_output_p99_ms": 5185.823632054962,
    "time_to_first_output_samples": 20,
    "time_to_first_output_stddev_ms": 38.75580740816014,
    "tpot_max_ms": 15.09121736212458,
    "tpot_mean_ci95_high_ms": 14.577620888727921,
    "tpot_mean_ci95_low_ms": 13.870427494566236,
    "tpot_mean_ms": 14.224024191647079,
    "tpot_min_ms": 12.958784748348053,
    "tpot_p25_ms": 13.068204114805102,
    "tpot_p50_ms": 14.67719238018617,
    "tpot_p75_ms": 14.708007530711223,
    "tpot_p90_ms": 14.739291469861822,
    "tpot_p95_ms": 14.830529432796505,
    "tpot_p99_ms": 15.09121736212458,
    "tpot_samples": 20,
    "tpot_stddev_ms": 0.8068022972369044,
    "ttft_max_ms": 222068.7335890252,
    "ttft_mean_ci95_high_ms": 48330.07958272229,
    "ttft_mean_ci95_low_ms": 1698.1346108672333,
    "ttft_mean_ms": 25014.10709679476,
    "ttft_min_ms": 5307.835966930725,
    "ttft_p25_ms": 6552.682909998111,
    "ttft_p50_ms": 7795.516223064624,
    "ttft_p75_ms": 8957.93227199465,
    "ttft_p90_ms": 23120.524459984154,
    "ttft_p95_ms": 124013.32262996584,
    "ttft_p99_ms": 222068.7335890252,
    "ttft_samples": 20,
    "ttft_stddev_ms": 53200.10146943746
  },
  "model": "llm.primary",
  "performance_eligible": true,
  "prompt_cache_mode": "unique",
  "request_canaries": false,
  "request_timings": [
    {
      "content_chunks": 40,
      "controlled_output": null,
      "decode_tok_s": 67.42847613980534,
      "e2e_ms": 5914.998011896387,
      "effective_prefill_tok_s": 4929.050266680819,
      "generation_ms": 786.0180599382147,
      "mean_inter_token_latency_ms": 14.830529432796505,
      "mean_time_per_output_token_ms": 14.830529432796505,
      "output_token_source": "usage",
      "output_tokens": 54,
      "planned_context_tokens": 32768,
      "prompt_tokens": 25281,
      "reasoning_chunks": 12,
      "request_canary": null,
      "request_index": 0,
      "time_to_first_output_ms": 5128.979951958172,
      "tpot_ms": 14.830529432796505,
      "ttft_ms": 5307.835966930725,
      "visible_content_truncated": false,
      "visible_generation_ms": 607.1620449656621
    },
    {
      "content_chunks": 59,
      "controlled_output": null,
      "decode_tok_s": 68.2035419225498,
      "e2e_ms": 23994.117031921633,
      "effective_prefill_tok_s": 4891.918746053634,
      "generation_ms": 18826.001756009646,
      "mean_inter_token_latency_ms": 14.661995137079163,
      "mean_time_per_output_token_ms": 14.661995137079163,
      "output_token_source": "usage",
      "output_tokens": 1285,
      "planned_context_tokens": 32768,
      "prompt_tokens": 25282,
      "reasoning_chunks": 1203,
      "request_canary": null,
      "request_index": 1,
      "time_to_first_output_ms": 5168.115275911987,
      "tpot_ms": 14.661995137079163,
      "ttft_ms": 23120.524459984154,
      "visible_content_truncated": false,
      "visible_generation_ms": 873.5925719374791
    },
    {
      "content_chunks": 64,
      "controlled_output": null,
      "decode_tok_s": 68.01315873486135,
      "e2e_ms": 9274.656245019287,
      "effective_prefill_tok_s": 4915.7096334788175,
      "generation_ms": 4131.553440936841,
      "mean_inter_token_latency_ms": 14.703037156358866,
      "mean_time_per_output_token_ms": 14.703037156358866,
      "output_token_source": "usage",
      "output_tokens": 282,
      "planned_context_tokens": 32768,
      "prompt_tokens": 25282,
      "reasoning_chunks": 210,
      "request_canary": null,
      "request_index": 2,
      "time_to_first_output_ms": 5143.102804082446,
      "tpot_ms": 14.703037156358866,
      "ttft_ms": 8312.998624052852,
      "visible_content_truncated": false,
      "visible_generation_ms": 961.6576209664345
    },
    {
      "content_chunks": 52,
      "controlled_output": null,
      "decode_tok_s": 67.99391545945338,
      "e2e_ms": 7980.754598043859,
      "effective_prefill_tok_s": 4916.31575317462,
      "generation_ms": 2838.4892779868096,
      "mean_inter_token_latency_ms": 14.707198331537874,
      "mean_time_per_output_token_ms": 14.707198331537874,
      "output_token_source": "usage",
      "output_tokens": 194,
      "planned_context_tokens": 32768,
      "prompt_tokens": 25281,
      "reasoning_chunks": 134,
      "request_canary": null,
      "request_index": 3,
      "time_to_first_output_ms": 5142.265320057049,
      "tpot_ms": 14.707198331537874,
      "ttft_ms": 7197.448487044312,
      "visible_content_truncated": false,
      "visible_generation_ms": 783.306110999547
    },
    {
      "content_chunks": 66,
      "controlled_output": null,
      "decode_tok_s": 66.26370663176357,
      "e2e_ms": 223057.7286890475,
      "effective_prefill_tok_s": 4875.214005298059,
      "generation_ms": 217871.90505699255,
      "mean_inter_token_latency_ms": 15.09121736212458,
      "mean_time_per_output_token_ms": 15.09121736212458,
      "output_token_source": "usage",
      "output_tokens": 14438,
      "planned_context_tokens": 32768,
      "prompt_tokens": 25282,
      "reasoning_chunks": 12472,
      "request_canary": null,
      "request_index": 4,
      "time_to_first_output_ms": 5185.823632054962,
      "tpot_ms": 15.09121736212458,
      "ttft_ms": 222068.7335890252,
      "visible_content_truncated": false,
      "visible_generation_ms": 988.9951000222936
    },
    {
      "content_chunks": 48,
      "controlled_output": null,
      "decode_tok_s": 68.13292175348019,
      "e2e_ms": 6619.133002939634,
      "effective_prefill_tok_s": 4907.973064048301,
      "generation_ms": 1467.719238018617,
      "mean_inter_token_latency_ms": 14.67719238018617,
      "mean_time_per_output_token_ms": 14.67719238018617,
      "output_token_source": "usage",
      "output_tokens": 101,
      "planned_context_tokens": 32768,
      "prompt_tokens": 25283,
      "reasoning_chunks": 50,
      "request_canary": null,
      "request_index": 5,
      "time_to_first_output_ms": 5151.413764921017,
      "tpot_ms": 14.67719238018617,
      "ttft_ms": 5904.726105974987,
      "visible_content_truncated": false,
      "visible_generation_ms": 714.4068969646469
    },
    {
      "content_chunks": 63,
      "controlled_output": null,
      "decode_tok_s": 67.99017459788068,
      "e2e_ms": 11411.616887082346,
      "effective_prefill_tok_s": 4885.399006039532,
      "generation_ms": 6236.195193021558,
      "mean_inter_token_latency_ms": 14.708007530711223,
      "mean_time_per_output_token_ms": 14.708007530711223,
      "output_token_source": "usage",
      "output_tokens": 425,
      "planned_context_tokens": 32768,
      "prompt_tokens": 25284,
      "reasoning_chunks": 354,
      "request_canary": null,
      "request_index": 6,
      "time_to_first_output_ms": 5175.421694060788,
      "tpot_ms": 14.708007530711223,
      "ttft_ms": 10460.637661046349,
      "visible_content_truncated": false,
      "visible_generation_ms": 950.9792260359973
    },
    {
      "content_chunks": 61,
      "controlled_output": null,
      "decode_tok_s": 68.02611750752207,
      "e2e_ms": 13492.34226893168,
      "effective_prefill_tok_s": 4902.751302574798,
      "generation_ms": 8335.033965995535,
      "mean_inter_token_latency_ms": 14.700236271597063,
      "mean_time_per_output_token_ms": 14.700236271597063,
      "output_token_source": "usage",
      "output_tokens": 568,
      "planned_context_tokens": 32768,
      "prompt_tokens": 25285,
      "reasoning_chunks": 492,
      "request_canary": null,
      "request_index": 7,
      "time_to_first_output_ms": 5157.308302936144,
      "tpot_ms": 14.700236271597063,
      "ttft_ms": 12579.414317966439,
      "visible_content_truncated": false,
      "visible_generation_ms": 912.9279509652406
    },
    {
      "content_chunks": 62,
      "controlled_output": null,
      "decode_tok_s": 67.86364254500008,
      "e2e_ms": 8612.800211063586,
      "effective_prefill_tok_s": 4923.821872220266,
      "generation_ms": 3477.561639039777,
      "mean_inter_token_latency_ms": 14.735430673897362,
      "mean_time_per_output_token_ms": 14.735430673897362,
      "output_token_source": "usage",
      "output_tokens": 237,
      "planned_context_tokens": 32768,
      "prompt_tokens": 25285,
      "reasoning_chunks": 169,
      "request_canary": null,
      "request_index": 8,
      "time_to_first_output_ms": 5135.238572023809,
      "tpot_ms": 14.735430673897362,
      "ttft_ms": 7683.347700047307,
      "visible_content_truncated": false,
      "visible_generation_ms": 929.4525110162795
    },
    {
      "content_chunks": 54,
      "controlled_output": null,
      "decode_tok_s": 67.84586640713027,
      "e2e_ms": 7594.8775940341875,
      "effective_prefill_tok_s": 4883.504900307471,
      "generation_ms": 2417.2438010573387,
      "mean_inter_token_latency_ms": 14.739291469861822,
      "mean_time_per_output_token_ms": 14.739291469861822,
      "output_token_source": "usage",
      "output_tokens": 165,
      "planned_context_tokens": 32768,
      "prompt_tokens": 25285,
      "reasoning_chunks": 105,
      "request_canary": null,
      "request_index": 9,
      "time_to_first_output_ms": 5177.633792976849,
      "tpot_ms": 14.739291469861822,
      "ttft_ms": 6783.968703937717,
      "visible_content_truncated": false,
      "visible_generation_ms": 810.9088900964707
    },
    {
      "content_chunks": 57,
      "controlled_output": null,
      "decode_tok_s": 68.23980522044606,
      "e2e_ms": 7301.461171009578,
      "effective_prefill_tok_s": 4898.5388314178235,
      "generation_ms": 2139.5137270446867,
      "mean_inter_token_latency_ms": 14.654203609895115,
      "mean_time_per_output_token_ms": 14.654203609895115,
      "output_token_source": "usage",
      "output_tokens": 147,
      "planned_context_tokens": 32768,
      "prompt_tokens": 25286,
      "reasoning_chunks": 87,
      "request_canary": null,
      "request_index": 10,
      "time_to_first_output_ms": 5161.947443964891,
      "tpot_ms": 14.654203609895115,
      "ttft_ms": 6458.746807998978,
      "visible_content_truncated": false,
      "visible_generation_ms": 842.7143630106002
    },
    {
      "content_chunks": 56,
      "controlled_output": null,
      "decode_tok_s": 68.19146259109985,
      "e2e_ms": 7665.755838970654,
      "effective_prefill_tok_s": 4930.410989473405,
      "generation_ms": 2536.9744749041274,
      "mean_inter_token_latency_ms": 14.664592340486285,
      "mean_time_per_output_token_ms": 14.664592340486285,
      "output_token_source": "usage",
      "output_tokens": 174,
      "planned_context_tokens": 32768,
      "prompt_tokens": 25287,
      "reasoning_chunks": 110,
      "request_canary": null,
      "request_index": 11,
      "time_to_first_output_ms": 5128.781364066526,
      "tpot_ms": 14.664592340486285,
      "ttft_ms": 6831.327782012522,
      "visible_content_truncated": false,
      "visible_generation_ms": 834.4280569581315
    },
    {
      "content_chunks": 63,
      "controlled_output": null,
      "decode_tok_s": 68.11678552574685,
      "e2e_ms": 8807.265256997198,
      "effective_prefill_tok_s": 4908.401917013103,
      "generation_ms": 3655.4866480873898,
      "mean_inter_token_latency_ms": 14.680669269427268,
      "mean_time_per_output_token_ms": 14.680669269427268,
      "output_token_source": "usage",
      "output_tokens": 250,
      "planned_context_tokens": 32768,
      "prompt_tokens": 25287,
      "reasoning_chunks": 182,
      "request_canary": null,
      "request_index": 12,
      "time_to_first_output_ms": 5151.778608909808,
      "tpot_ms": 14.680669269427268,
      "ttft_ms": 7867.342379991896,
      "visible_content_truncated": false,
      "visible_generation_ms": 939.9228770053014
    },
    {
      "content_chunks": 50,
      "controlled_output": null,
      "decode_tok_s": 67.86795138092191,
      "e2e_ms": 8548.55074500665,
      "effective_prefill_tok_s": 4900.751485224337,
      "generation_ms": 3388.9338829321787,
      "mean_inter_token_latency_ms": 14.734495143183386,
      "mean_time_per_output_token_ms": 14.734495143183386,
      "output_token_source": "usage",
      "output_tokens": 231,
      "planned_context_tokens": 32768,
      "prompt_tokens": 25286,
      "reasoning_chunks": 166,
      "request_canary": null,
      "request_index": 13,
      "time_to_first_output_ms": 5159.616862074472,
      "tpot_ms": 14.734495143183386,
      "ttft_ms": 7795.516223064624,
      "visible_content_truncated": false,
      "visible_generation_ms": 753.0345219420269
    },
    {
      "content_chunks": 65,
      "controlled_output": null,
      "decode_tok_s": 75.97270633104887,
      "e2e_ms": 124872.20033397898,
      "effective_prefill_tok_s": 4914.681597827631,
      "generation_ms": 119727.2078259848,
      "mean_inter_token_latency_ms": 13.162621792654441,
      "mean_time_per_output_token_ms": 13.162621792654441,
      "output_token_source": "usage",
      "output_tokens": 9097,
      "planned_context_tokens": 32768,
      "prompt_tokens": 25286,
      "reasoning_chunks": 8670,
      "request_canary": null,
      "request_index": 14,
      "time_to_first_output_ms": 5144.992507994175,
      "tpot_ms": 13.162621792654441,
      "ttft_ms": 124013.32262996584,
      "visible_content_truncated": false,
      "visible_generation_ms": 858.877704013139
    },
    {
      "content_chunks": 55,
      "controlled_output": null,
      "decode_tok_s": 77.16772980024048,
      "e2e_ms": 7274.534767027944,
      "effective_prefill_tok_s": 4998.633436303636,
      "generation_ms": 2215.952191967517,
      "mean_inter_token_latency_ms": 12.958784748348053,
      "mean_time_per_output_token_ms": 12.958784748348053,
      "output_token_source": "usage",
      "output_tokens": 172,
      "planned_context_tokens": 32768,
      "prompt_tokens": 25286,
      "reasoning_chunks": 111,
      "request_canary": null,
      "request_index": 15,
      "time_to_first_output_ms": 5058.582575060427,
      "tpot_ms": 12.958784748348053,
      "ttft_ms": 6552.682909998111,
      "visible_content_truncated": false,
      "visible_generation_ms": 721.8518570298329
    },
    {
      "content_chunks": 61,
      "controlled_output": null,
      "decode_tok_s": 76.52160857107289,
      "e2e_ms": 6987.902516964823,
      "effective_prefill_tok_s": 5003.553450989291,
      "generation_ms": 1934.094208991155,
      "mean_inter_token_latency_ms": 13.068204114805102,
      "mean_time_per_output_token_ms": 13.068204114805102,
      "output_token_source": "usage",
      "output_tokens": 149,
      "planned_context_tokens": 32768,
      "prompt_tokens": 25287,
      "reasoning_chunks": 83,
      "request_canary": null,
      "request_index": 16,
      "time_to_first_output_ms": 5053.808307973668,
      "tpot_ms": 13.068204114805102,
      "ttft_ms": 6175.656238920055,
      "visible_content_truncated": false,
      "visible_generation_ms": 812.2462780447677
    },
    {
      "content_chunks": 81,
      "controlled_output": null,
      "decode_tok_s": 76.86001228799726,
      "e2e_ms": 10031.399628031068,
      "effective_prefill_tok_s": 4970.371297394294,
      "generation_ms": 4944.05333395116,
      "mean_inter_token_latency_ms": 13.010666668292528,
      "mean_time_per_output_token_ms": 13.010666668292528,
      "output_token_source": "usage",
      "output_tokens": 381,
      "planned_context_tokens": 32768,
      "prompt_tokens": 25286,
      "reasoning_chunks": 295,
      "request_canary": null,
      "request_index": 17,
      "time_to_first_output_ms": 5087.346294079907,
      "tpot_ms": 13.010666668292528,
      "ttft_ms": 8957.93227199465,
      "visible_content_truncated": false,
      "visible_generation_ms": 1073.4673560364172
    },
    {
      "content_chunks": 80,
      "controlled_output": null,
      "decode_tok_s": 76.93138051397554,
      "e2e_ms": 9346.874031005427,
      "effective_prefill_tok_s": 4974.490890606123,
      "generation_ms": 4263.539765030146,
      "mean_inter_token_latency_ms": 12.998596844604103,
      "mean_time_per_output_token_ms": 12.998596844604103,
      "output_token_source": "usage",
      "output_tokens": 329,
      "planned_context_tokens": 32768,
      "prompt_tokens": 25287,
      "reasoning_chunks": 240,
      "request_canary": null,
      "request_index": 18,
      "time_to_first_output_ms": 5083.334265975282,
      "tpot_ms": 12.998596844604103,
      "ttft_ms": 8291.929811937734,
      "visible_content_truncated": false,
      "visible_generation_ms": 1054.9442190676928
    },
    {
      "content_chunks": 81,
      "controlled_output": null,
      "decode_tok_s": 76.96147741408313,
      "e2e_ms": 8983.139773015864,
      "effective_prefill_tok_s": 4960.299538026573,
      "generation_ms": 3885.0605529733,
      "mean_inter_token_latency_ms": 12.99351355509465,
      "mean_time_per_output_token_ms": 12.99351355509465,
      "output_token_source": "usage",
      "output_tokens": 300,
      "planned_context_tokens": 32768,
      "prompt_tokens": 25288,
      "reasoning_chunks": 214,
      "request_canary": null,
      "request_index": 19,
      "time_to_first_output_ms": 5098.079220042564,
      "tpot_ms": 12.99351355509465,
      "ttft_ms": 7918.049264000729,
      "visible_content_truncated": false,
      "visible_generation_ms": 1065.0905090151355
    }
  ],
  "requests": 20,
  "response_words": 0,
  "responses_received": 20,
  "run_id": "benchmark-20261017T002620Z",
  "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": "2026-10-17T00:26:20Z",
  "started_at_unix_ns": 1792196780313055235,
  "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": 517806.1951720156
}