{
  "base_url": "redacted://routed-capability",
  "clock_domain_id": null,
  "completed": 20,
  "concurrency": 1,
  "configuration_fingerprint": null,
  "context_distribution": {
    "32768": 20
  },
  "context_seed": 941321,
  "context_tokens": 32768,
  "controlled_output_policy": "observe",
  "engine": "vllm-exl3-v84",
  "failed": 0,
  "failures": [],
  "finished_at": "2026-10-17T00:50:17Z",
  "finished_at_unix_ns": 1792198217517961102,
  "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": 563.0,
    "completion_tokens_mean": 220.1,
    "completion_tokens_mean_ci95_high": 277.6020288081441,
    "completion_tokens_mean_ci95_low": 162.59797119185592,
    "completion_tokens_min": 87.0,
    "completion_tokens_p25": 108.0,
    "completion_tokens_p50": 204.0,
    "completion_tokens_p75": 256.0,
    "completion_tokens_p90": 395.0,
    "completion_tokens_p95": 463.0,
    "completion_tokens_p99": 563.0,
    "completion_tokens_samples": 20,
    "completion_tokens_stddev": 131.20249516240958,
    "content_chunks": 1100,
    "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": 77.47432379043512,
    "decode_tok_s_mean": 76.89191793179216,
    "decode_tok_s_mean_ci95_high": 77.00645536974372,
    "decode_tok_s_mean_ci95_low": 76.77738049384061,
    "decode_tok_s_min": 76.52391957948298,
    "decode_tok_s_p25": 76.59525539209979,
    "decode_tok_s_p50": 76.91591228174815,
    "decode_tok_s_p75": 77.04734804124436,
    "decode_tok_s_p90": 77.15179586507242,
    "decode_tok_s_p95": 77.16864916962831,
    "decode_tok_s_p99": 77.47432379043512,
    "decode_tok_s_samples": 20,
    "decode_tok_s_stddev": 0.26134030329423147,
    "e2e_max_ms": 12283.423815038987,
    "e2e_mean_ci95_high_ms": 8645.21090155845,
    "e2e_mean_ci95_low_ms": 7163.066970925727,
    "e2e_mean_ms": 7904.138936242089,
    "e2e_min_ms": 6188.482488971204,
    "e2e_p25_ms": 6479.302118998021,
    "e2e_p50_ms": 7676.636946969666,
    "e2e_p75_ms": 8339.836576022208,
    "e2e_p90_ms": 10206.82849700097,
    "e2e_p95_ms": 11074.38229105901,
    "e2e_p99_ms": 12283.423815038987,
    "e2e_samples": 20,
    "e2e_stddev_ms": 1690.905398639542,
    "effective_prefill_tok_s_max": 5144.869960097099,
    "effective_prefill_tok_s_mean": 5001.9071461259955,
    "effective_prefill_tok_s_mean_ci95_high": 5020.727377780919,
    "effective_prefill_tok_s_mean_ci95_low": 4983.086914471072,
    "effective_prefill_tok_s_min": 4949.020423813218,
    "effective_prefill_tok_s_p25": 4980.770876021492,
    "effective_prefill_tok_s_p50": 4989.322004308864,
    "effective_prefill_tok_s_p75": 5004.237694712734,
    "effective_prefill_tok_s_p90": 5037.213478031881,
    "effective_prefill_tok_s_p95": 5078.458409972194,
    "effective_prefill_tok_s_p99": 5144.869960097099,
    "effective_prefill_tok_s_samples": 20,
    "effective_prefill_tok_s_stddev": 42.94216054357341,
    "generation_max_ms": 7305.141438962892,
    "generation_mean_ci95_high_ms": 3595.950092189054,
    "generation_mean_ci95_low_ms": 2101.563742477407,
    "generation_mean_ms": 2848.7569173332304,
    "generation_min_ms": 1114.4422109937295,
    "generation_p25_ms": 1396.9533680938184,
    "generation_p50_ms": 2631.1766009312123,
    "generation_p75_ms": 3320.9903209935874,
    "generation_p90_ms": 5123.55653999839,
    "generation_p95_ms": 6006.559452973306,
    "generation_p99_ms": 7305.141438962892,
    "generation_samples": 20,
    "generation_stddev_ms": 1704.8721747974573,
    "mean_inter_token_latency_ms_max": 13.067809457425028,
    "mean_inter_token_latency_ms_mean": 13.005410665128494,
    "mean_inter_token_latency_ms_mean_ci95_high": 13.024774996525476,
    "mean_inter_token_latency_ms_mean_ci95_low": 12.986046333731512,
    "mean_inter_token_latency_ms_min": 12.907502138449882,
    "mean_inter_token_latency_ms_p25": 12.976957348003705,
    "mean_inter_token_latency_ms_p50": 12.998472311321871,
    "mean_inter_token_latency_ms_p75": 13.05513955713784,
    "mean_inter_token_latency_ms_p90": 13.06459341209293,
    "mean_inter_token_latency_ms_p95": 13.065552240706557,
    "mean_inter_token_latency_ms_p99": 13.067809457425028,
    "mean_inter_token_latency_ms_samples": 20,
    "mean_inter_token_latency_ms_stddev": 0.04418363402294494,
    "mean_time_per_output_token_max_ms": 13.067809457425028,
    "mean_time_per_output_token_mean_ci95_high_ms": 13.024774996525476,
    "mean_time_per_output_token_mean_ci95_low_ms": 12.986046333731512,
    "mean_time_per_output_token_mean_ms": 13.005410665128494,
    "mean_time_per_output_token_min_ms": 12.907502138449882,
    "mean_time_per_output_token_p25_ms": 12.976957348003705,
    "mean_time_per_output_token_p50_ms": 12.998472311321871,
    "mean_time_per_output_token_p75_ms": 13.05513955713784,
    "mean_time_per_output_token_p90_ms": 13.06459341209293,
    "mean_time_per_output_token_p95_ms": 13.065552240706557,
    "mean_time_per_output_token_p99_ms": 13.067809457425028,
    "mean_time_per_output_token_samples": 20,
    "mean_time_per_output_token_stddev_ms": 0.04418363402294494,
    "output_token_sources": {
      "usage": 20
    },
    "output_tokens": 4402,
    "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": 3190,
    "request_canary_passed": 0,
    "request_canary_samples": 0,
    "throughput_performance_eligible": true,
    "throughput_tok_s": 27.84095090307982,
    "time_to_first_output_max_ms": 5109.495987999253,
    "time_to_first_output_mean_ci95_high_ms": 5074.200379776201,
    "time_to_first_output_mean_ci95_low_ms": 5036.563658041516,
    "time_to_first_output_mean_ms": 5055.382018908858,
    "time_to_first_output_min_ms": 4913.826820906252,
    "time_to_first_output_p25_ms": 5052.256005001254,
    "time_to_first_output_p50_ms": 5065.331352059729,
    "time_to_first_output_p75_ms": 5076.908451970667,
    "time_to_first_output_p90_ms": 5083.27195700258,
    "time_to_first_output_p95_ms": 5086.933783022687,
    "time_to_first_output_p99_ms": 5109.495987999253,
    "time_to_first_output_samples": 20,
    "time_to_first_output_stddev_ms": 42.93789196377461,
    "tpot_max_ms": 13.067809457425028,
    "tpot_mean_ci95_high_ms": 13.024774996525476,
    "tpot_mean_ci95_low_ms": 12.986046333731512,
    "tpot_mean_ms": 13.005410665128494,
    "tpot_min_ms": 12.907502138449882,
    "tpot_p25_ms": 12.976957348003705,
    "tpot_p50_ms": 12.998472311321871,
    "tpot_p75_ms": 13.05513955713784,
    "tpot_p90_ms": 13.06459341209293,
    "tpot_p95_ms": 13.065552240706557,
    "tpot_p99_ms": 13.067809457425028,
    "tpot_samples": 20,
    "tpot_stddev_ms": 0.04418363402294494,
    "ttft_max_ms": 11013.720307033509,
    "ttft_mean_ci95_high_ms": 7862.480598885376,
    "ttft_mean_ci95_low_ms": 6485.743558023315,
    "ttft_mean_ms": 7174.1120784543455,
    "ttft_min_ms": 5551.893098978326,
    "ttft_p25_ms": 5890.723765012808,
    "ttft_p50_ms": 7045.820003957488,
    "ttft_p75_ms": 7644.502833951265,
    "ttft_p90_ms": 9512.067199102603,
    "ttft_p95_ms": 10092.888249084353,
    "ttft_p99_ms": 11013.720307033509,
    "ttft_samples": 20,
    "ttft_stddev_ms": 1570.6518420966697
  },
  "model": "llm.primary",
  "performance_eligible": true,
  "prompt_cache_mode": "unique",
  "request_canaries": false,
  "request_timings": [
    {
      "content_chunks": 52,
      "controlled_output": null,
      "decode_tok_s": 76.99533241486121,
      "e2e_ms": 7823.094083927572,
      "effective_prefill_tok_s": 5144.869960097099,
      "generation_ms": 2909.26726302132,
      "mean_inter_token_latency_ms": 12.987800281345178,
      "mean_time_per_output_token_ms": 12.987800281345178,
      "output_token_source": "usage",
      "output_tokens": 225,
      "planned_context_tokens": 32768,
      "prompt_tokens": 25281,
      "reasoning_chunks": 169,
      "request_canary": null,
      "request_index": 0,
      "time_to_first_output_ms": 4913.826820906252,
      "tpot_ms": 12.987800281345178,
      "ttft_ms": 7137.083351961337,
      "visible_content_truncated": false,
      "visible_generation_ms": 686.0107319662347
    },
    {
      "content_chunks": 96,
      "controlled_output": null,
      "decode_tok_s": 76.93211756346048,
      "e2e_ms": 12283.423815038987,
      "effective_prefill_tok_s": 5078.458409972194,
      "generation_ms": 7305.141438962892,
      "mean_inter_token_latency_ms": 12.998472311321871,
      "mean_time_per_output_token_ms": 12.998472311321871,
      "output_token_source": "usage",
      "output_tokens": 563,
      "planned_context_tokens": 32768,
      "prompt_tokens": 25282,
      "reasoning_chunks": 458,
      "request_canary": null,
      "request_index": 1,
      "time_to_first_output_ms": 4978.282376076095,
      "tpot_ms": 12.998472311321871,
      "ttft_ms": 11013.720307033509,
      "visible_content_truncated": false,
      "visible_generation_ms": 1269.7035080054775
    },
    {
      "content_chunks": 48,
      "controlled_output": null,
      "decode_tok_s": 77.15179586507242,
      "e2e_ms": 7676.636946969666,
      "effective_prefill_tok_s": 5010.841086056831,
      "generation_ms": 2631.1766009312123,
      "mean_inter_token_latency_ms": 12.961461088331095,
      "mean_time_per_output_token_ms": 12.961461088331095,
      "output_token_source": "usage",
      "output_tokens": 204,
      "planned_context_tokens": 32768,
      "prompt_tokens": 25282,
      "reasoning_chunks": 148,
      "request_canary": null,
      "request_index": 2,
      "time_to_first_output_ms": 5045.460346038453,
      "tpot_ms": 12.961461088331095,
      "ttft_ms": 7045.820003957488,
      "visible_content_truncated": false,
      "visible_generation_ms": 630.816943012178
    },
    {
      "content_chunks": 59,
      "controlled_output": null,
      "decode_tok_s": 76.78432496114836,
      "e2e_ms": 8339.836576022208,
      "effective_prefill_tok_s": 5037.213478031881,
      "generation_ms": 3320.9903209935874,
      "mean_inter_token_latency_ms": 13.023491454876813,
      "mean_time_per_output_token_ms": 13.023491454876813,
      "output_token_source": "usage",
      "output_tokens": 256,
      "planned_context_tokens": 32768,
      "prompt_tokens": 25281,
      "reasoning_chunks": 187,
      "request_canary": null,
      "request_index": 3,
      "time_to_first_output_ms": 5018.84625502862,
      "tpot_ms": 13.023491454876813,
      "ttft_ms": 7550.350047065876,
      "visible_content_truncated": false,
      "visible_generation_ms": 789.4865289563313
    },
    {
      "content_chunks": 52,
      "controlled_output": null,
      "decode_tok_s": 76.8997076394367,
      "e2e_ms": 10206.82849700097,
      "effective_prefill_tok_s": 4973.568247744878,
      "generation_ms": 5123.55653999839,
      "mean_inter_token_latency_ms": 13.00395060913297,
      "mean_time_per_output_token_ms": 13.00395060913297,
      "output_token_source": "usage",
      "output_tokens": 395,
      "planned_context_tokens": 32768,
      "prompt_tokens": 25282,
      "reasoning_chunks": 332,
      "request_canary": null,
      "request_index": 4,
      "time_to_first_output_ms": 5083.27195700258,
      "tpot_ms": 13.00395060913297,
      "ttft_ms": 9512.067199102603,
      "visible_content_truncated": false,
      "visible_generation_ms": 694.761297898367
    },
    {
      "content_chunks": 47,
      "controlled_output": null,
      "decode_tok_s": 76.87577759816033,
      "e2e_ms": 6340.047856909223,
      "effective_prefill_tok_s": 5004.299064610389,
      "generation_ms": 1287.7918519079685,
      "mean_inter_token_latency_ms": 13.007998504120895,
      "mean_time_per_output_token_ms": 13.007998504120895,
      "output_token_source": "usage",
      "output_tokens": 100,
      "planned_context_tokens": 32768,
      "prompt_tokens": 25283,
      "reasoning_chunks": 49,
      "request_canary": null,
      "request_index": 5,
      "time_to_first_output_ms": 5052.256005001254,
      "tpot_ms": 13.007998504120895,
      "ttft_ms": 5719.408822944388,
      "visible_content_truncated": false,
      "visible_generation_ms": 620.6390339648351
    },
    {
      "content_chunks": 55,
      "controlled_output": null,
      "decode_tok_s": 77.47432379043512,
      "e2e_ms": 6741.598427994177,
      "effective_prefill_tok_s": 4993.262581166118,
      "generation_ms": 1677.9752779984847,
      "mean_inter_token_latency_ms": 12.907502138449882,
      "mean_time_per_output_token_ms": 12.907502138449882,
      "output_token_source": "usage",
      "output_tokens": 131,
      "planned_context_tokens": 32768,
      "prompt_tokens": 25284,
      "reasoning_chunks": 74,
      "request_canary": null,
      "request_index": 6,
      "time_to_first_output_ms": 5063.623149995692,
      "tpot_ms": 12.907502138449882,
      "ttft_ms": 6021.114932955243,
      "visible_content_truncated": false,
      "visible_generation_ms": 720.483495038934
    },
    {
      "content_chunks": 59,
      "controlled_output": null,
      "decode_tok_s": 76.54275708835866,
      "e2e_ms": 6698.85639497079,
      "effective_prefill_tok_s": 5004.237694712734,
      "generation_ms": 1646.1387699237093,
      "mean_inter_token_latency_ms": 13.06459341209293,
      "mean_time_per_output_token_ms": 13.06459341209293,
      "output_token_source": "usage",
      "output_tokens": 127,
      "planned_context_tokens": 32768,
      "prompt_tokens": 25285,
      "reasoning_chunks": 64,
      "request_canary": null,
      "request_index": 7,
      "time_to_first_output_ms": 5052.71762504708,
      "tpot_ms": 13.06459341209293,
      "ttft_ms": 5915.1204290101305,
      "visible_content_truncated": false,
      "visible_generation_ms": 783.7359659606591
    },
    {
      "content_chunks": 74,
      "controlled_output": null,
      "decode_tok_s": 76.91591228174815,
      "e2e_ms": 11074.38229105901,
      "effective_prefill_tok_s": 4989.322004308864,
      "generation_ms": 6006.559452973306,
      "mean_inter_token_latency_ms": 13.001210937171658,
      "mean_time_per_output_token_ms": 13.001210937171658,
      "output_token_source": "usage",
      "output_tokens": 463,
      "planned_context_tokens": 32768,
      "prompt_tokens": 25285,
      "reasoning_chunks": 375,
      "request_canary": null,
      "request_index": 8,
      "time_to_first_output_ms": 5067.822838085704,
      "tpot_ms": 13.001210937171658,
      "ttft_ms": 10092.888249084353,
      "visible_content_truncated": false,
      "visible_generation_ms": 981.4940419746563
    },
    {
      "content_chunks": 48,
      "controlled_output": null,
      "decode_tok_s": 77.16864916962831,
      "e2e_ms": 6188.482488971204,
      "effective_prefill_tok_s": 4983.208373363301,
      "generation_ms": 1114.4422109937295,
      "mean_inter_token_latency_ms": 12.958630360392204,
      "mean_time_per_output_token_ms": 12.958630360392204,
      "output_token_source": "usage",
      "output_tokens": 87,
      "planned_context_tokens": 32768,
      "prompt_tokens": 25285,
      "reasoning_chunks": 36,
      "request_canary": null,
      "request_index": 9,
      "time_to_first_output_ms": 5074.040277977474,
      "tpot_ms": 12.958630360392204,
      "ttft_ms": 5551.893098978326,
      "visible_content_truncated": false,
      "visible_generation_ms": 636.5893899928778
    },
    {
      "content_chunks": 44,
      "controlled_output": null,
      "decode_tok_s": 76.5981853831087,
      "e2e_ms": 8241.881250985898,
      "effective_prefill_tok_s": 5000.764176883513,
      "generation_ms": 3185.454051941633,
      "mean_inter_token_latency_ms": 13.05513955713784,
      "mean_time_per_output_token_ms": 13.05513955713784,
      "output_token_source": "usage",
      "output_tokens": 245,
      "planned_context_tokens": 32768,
      "prompt_tokens": 25286,
      "reasoning_chunks": 197,
      "request_canary": null,
      "request_index": 10,
      "time_to_first_output_ms": 5056.427199044265,
      "tpot_ms": 13.05513955713784,
      "ttft_ms": 7644.502833951265,
      "visible_content_truncated": false,
      "visible_generation_ms": 597.3784170346335
    },
    {
      "content_chunks": 55,
      "controlled_output": null,
      "decode_tok_s": 76.58708071090676,
      "e2e_ms": 6643.752348958515,
      "effective_prefill_tok_s": 4980.787075288807,
      "generation_ms": 1566.843896987848,
      "mean_inter_token_latency_ms": 13.057032474898733,
      "mean_time_per_output_token_ms": 13.057032474898733,
      "output_token_source": "usage",
      "output_tokens": 121,
      "planned_context_tokens": 32768,
      "prompt_tokens": 25287,
      "reasoning_chunks": 63,
      "request_canary": null,
      "request_index": 11,
      "time_to_first_output_ms": 5076.908451970667,
      "tpot_ms": 13.057032474898733,
      "ttft_ms": 5909.92492903024,
      "visible_content_truncated": false,
      "visible_generation_ms": 733.827419928275
    },
    {
      "content_chunks": 40,
      "controlled_output": null,
      "decode_tok_s": 77.13165467999885,
      "e2e_ms": 6369.306771084666,
      "effective_prefill_tok_s": 4984.7991882748,
      "generation_ms": 1296.4845680398867,
      "mean_inter_token_latency_ms": 12.964845680398867,
      "mean_time_per_output_token_ms": 12.964845680398867,
      "output_token_source": "usage",
      "output_tokens": 101,
      "planned_context_tokens": 32768,
      "prompt_tokens": 25287,
      "reasoning_chunks": 57,
      "request_canary": null,
      "request_index": 12,
      "time_to_first_output_ms": 5072.82220304478,
      "tpot_ms": 12.964845680398867,
      "ttft_ms": 5839.326846995391,
      "visible_content_truncated": false,
      "visible_generation_ms": 529.9799240892753
    },
    {
      "content_chunks": 44,
      "controlled_output": null,
      "decode_tok_s": 76.59525539209979,
      "e2e_ms": 6479.302118998021,
      "effective_prefill_tok_s": 4975.258731605413,
      "generation_ms": 1396.9533680938184,
      "mean_inter_token_latency_ms": 13.055638954147836,
      "mean_time_per_output_token_ms": 13.055638954147836,
      "output_token_source": "usage",
      "output_tokens": 108,
      "planned_context_tokens": 32768,
      "prompt_tokens": 25286,
      "reasoning_chunks": 60,
      "request_canary": null,
      "request_index": 13,
      "time_to_first_output_ms": 5082.3487509042025,
      "tpot_ms": 13.055638954147836,
      "ttft_ms": 5890.723765012808,
      "visible_content_truncated": false,
      "visible_generation_ms": 588.5783539852127
    },
    {
      "content_chunks": 45,
      "controlled_output": null,
      "decode_tok_s": 76.52391957948298,
      "e2e_ms": 6315.3078720206395,
      "effective_prefill_tok_s": 4970.774356133826,
      "generation_ms": 1228.3740889979526,
      "mean_inter_token_latency_ms": 13.067809457425028,
      "mean_time_per_output_token_ms": 13.067809457425028,
      "output_token_source": "usage",
      "output_tokens": 95,
      "planned_context_tokens": 32768,
      "prompt_tokens": 25286,
      "reasoning_chunks": 47,
      "request_canary": null,
      "request_index": 14,
      "time_to_first_output_ms": 5086.933783022687,
      "tpot_ms": 13.067809457425028,
      "ttft_ms": 5710.992943961173,
      "visible_content_truncated": false,
      "visible_generation_ms": 604.3149280594662
    },
    {
      "content_chunks": 51,
      "controlled_output": null,
      "decode_tok_s": 77.04120411029497,
      "e2e_ms": 7887.645930051804,
      "effective_prefill_tok_s": 4986.421372727184,
      "generation_ms": 2816.6745640337467,
      "mean_inter_token_latency_ms": 12.980067115362887,
      "mean_time_per_output_token_ms": 12.980067115362887,
      "output_token_source": "usage",
      "output_tokens": 218,
      "planned_context_tokens": 32768,
      "prompt_tokens": 25286,
      "reasoning_chunks": 159,
      "request_canary": null,
      "request_index": 15,
      "time_to_first_output_ms": 5070.971366018057,
      "tpot_ms": 12.980067115362887,
      "ttft_ms": 7211.333803017624,
      "visible_content_truncated": false,
      "visible_generation_ms": 676.3121270341799
    },
    {
      "content_chunks": 63,
      "controlled_output": null,
      "decode_tok_s": 76.53713992160519,
      "e2e_ms": 6951.738853938878,
      "effective_prefill_tok_s": 4949.020423813218,
      "generation_ms": 1842.2428659396246,
      "mean_inter_token_latency_ms": 13.065552240706557,
      "mean_time_per_output_token_ms": 13.065552240706557,
      "output_token_source": "usage",
      "output_tokens": 142,
      "planned_context_tokens": 32768,
      "prompt_tokens": 25287,
      "reasoning_chunks": 76,
      "request_canary": null,
      "request_index": 16,
      "time_to_first_output_ms": 5109.495987999253,
      "tpot_ms": 13.065552240706557,
      "ttft_ms": 6111.8004189338535,
      "visible_content_truncated": false,
      "visible_generation_ms": 839.9384350050241
    },
    {
      "content_chunks": 64,
      "controlled_output": null,
      "decode_tok_s": 77.05966608219097,
      "e2e_ms": 8867.579855024815,
      "effective_prefill_tok_s": 4991.973524045547,
      "generation_ms": 3802.248502965085,
      "mean_inter_token_latency_ms": 12.976957348003705,
      "mean_time_per_output_token_ms": 12.976957348003705,
      "output_token_source": "usage",
      "output_tokens": 294,
      "planned_context_tokens": 32768,
      "prompt_tokens": 25286,
      "reasoning_chunks": 224,
      "request_canary": null,
      "request_index": 17,
      "time_to_first_output_ms": 5065.331352059729,
      "tpot_ms": 12.976957348003705,
      "ttft_ms": 8025.302906055003,
      "visible_content_truncated": false,
      "visible_generation_ms": 842.2769489698112
    },
    {
      "content_chunks": 56,
      "controlled_output": null,
      "decode_tok_s": 76.97620636260103,
      "e2e_ms": 8125.010344898328,
      "effective_prefill_tok_s": 4998.292297661821,
      "generation_ms": 3065.882447990589,
      "mean_inter_token_latency_ms": 12.99102732199402,
      "mean_time_per_output_token_ms": 12.99102732199402,
      "output_token_source": "usage",
      "output_tokens": 237,
      "planned_context_tokens": 32768,
      "prompt_tokens": 25287,
      "reasoning_chunks": 178,
      "request_canary": null,
      "request_index": 18,
      "time_to_first_output_ms": 5059.127896907739,
      "tpot_ms": 12.99102732199402,
      "ttft_ms": 7388.207382988185,
      "visible_content_truncated": false,
      "visible_generation_ms": 736.8029619101435
    },
    {
      "content_chunks": 48,
      "controlled_output": null,
      "decode_tok_s": 77.04734804124436,
      "e2e_ms": 8828.066000016406,
      "effective_prefill_tok_s": 4980.770876021492,
      "generation_ms": 3750.9402639698237,
      "mean_inter_token_latency_ms": 12.979032055258905,
      "mean_time_per_output_token_ms": 12.979032055258905,
      "output_token_source": "usage",
      "output_tokens": 290,
      "planned_context_tokens": 32768,
      "prompt_tokens": 25288,
      "reasoning_chunks": 237,
      "request_canary": null,
      "request_index": 19,
      "time_to_first_output_ms": 5077.1257360465825,
      "tpot_ms": 12.979032055258905,
      "ttft_ms": 8190.659297048114,
      "visible_content_truncated": false,
      "visible_generation_ms": 637.4067029682919
    }
  ],
  "requests": 20,
  "response_words": 0,
  "responses_received": 20,
  "run_id": "benchmark-20261017T004739Z",
  "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": "2026-10-17T00:47:39Z",
  "started_at_unix_ns": 1792198059405540463,
  "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": 158112.41560406052
}