{
  "base_url": "redacted://routed-capability",
  "clock_domain_id": null,
  "completed": 20,
  "concurrency": 1,
  "configuration_fingerprint": null,
  "context_distribution": {
    "8192": 20
  },
  "context_seed": 941111,
  "context_tokens": 8192,
  "controlled_output_policy": "observe",
  "engine": "vllm-exl3-v84",
  "failed": 0,
  "failures": [],
  "finished_at": "2026-10-17T00:17:16Z",
  "finished_at_unix_ns": 1792196236944882636,
  "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": 1016.0,
    "completion_tokens_mean": 268.15,
    "completion_tokens_mean_ci95_high": 356.77356651175324,
    "completion_tokens_mean_ci95_low": 179.5264334882467,
    "completion_tokens_min": 57.0,
    "completion_tokens_p25": 145.0,
    "completion_tokens_p50": 211.0,
    "completion_tokens_p75": 324.0,
    "completion_tokens_p90": 378.0,
    "completion_tokens_p95": 422.0,
    "completion_tokens_p99": 1016.0,
    "completion_tokens_samples": 20,
    "completion_tokens_stddev": 202.21257053954517,
    "content_chunks": 969,
    "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": 77.10555821375755,
    "decode_tok_s_mean": 69.4152482331023,
    "decode_tok_s_mean_ci95_high": 71.19352961731545,
    "decode_tok_s_mean_ci95_low": 67.63696684888916,
    "decode_tok_s_min": 62.07052345115315,
    "decode_tok_s_p25": 67.7430379109125,
    "decode_tok_s_p50": 68.00951177144326,
    "decode_tok_s_p75": 68.15175800852141,
    "decode_tok_s_p90": 76.82391773448269,
    "decode_tok_s_p95": 76.91526678458256,
    "decode_tok_s_p99": 77.10555821375755,
    "decode_tok_s_samples": 20,
    "decode_tok_s_stddev": 4.057508222676536,
    "e2e_max_ms": 16230.59262696188,
    "e2e_mean_ci95_high_ms": 6445.855681595882,
    "e2e_mean_ci95_low_ms": 3860.021559209539,
    "e2e_mean_ms": 5152.9386204027105,
    "e2e_min_ms": 2166.525493026711,
    "e2e_p25_ms": 3480.660113040358,
    "e2e_p50_ms": 4120.3440389363095,
    "e2e_p75_ms": 6061.102248029783,
    "e2e_p90_ms": 6725.745756994002,
    "e2e_p95_ms": 6874.3226629449055,
    "e2e_p99_ms": 16230.59262696188,
    "e2e_samples": 20,
    "e2e_stddev_ms": 2950.051467446109,
    "effective_prefill_tok_s_max": 4857.362104511552,
    "effective_prefill_tok_s_mean": 4636.480585784563,
    "effective_prefill_tok_s_mean_ci95_high": 4688.7373518091035,
    "effective_prefill_tok_s_mean_ci95_low": 4584.223819760023,
    "effective_prefill_tok_s_min": 4487.457530880858,
    "effective_prefill_tok_s_p25": 4541.986156795802,
    "effective_prefill_tok_s_p50": 4592.396594816607,
    "effective_prefill_tok_s_p75": 4659.518706348461,
    "effective_prefill_tok_s_p90": 4818.952977169809,
    "effective_prefill_tok_s_p95": 4845.909899854215,
    "effective_prefill_tok_s_p99": 4857.362104511552,
    "effective_prefill_tok_s_samples": 20,
    "effective_prefill_tok_s_stddev": 119.23436848487052,
    "generation_max_ms": 14893.232833012007,
    "generation_mean_ci95_high_ms": 5134.42064272837,
    "generation_mean_ci95_low_ms": 2550.287939660333,
    "generation_mean_ms": 3842.3542911943514,
    "generation_min_ms": 828.5645679570735,
    "generation_p25_ms": 2137.4243369791657,
    "generation_p50_ms": 2814.8044199915603,
    "generation_p75_ms": 4743.762616068125,
    "generation_p90_ms": 5473.555739969015,
    "generation_p95_ms": 5543.34224993363,
    "generation_p99_ms": 14893.232833012007,
    "generation_samples": 20,
    "generation_stddev_ms": 2948.1104015002097,
    "mean_inter_token_latency_ms_max": 16.110706731625314,
    "mean_inter_token_latency_ms_mean": 14.450631174528056,
    "mean_inter_token_latency_ms_mean_ci95_high": 14.803233410672602,
    "mean_inter_token_latency_ms_mean_ci95_low": 14.098028938383509,
    "mean_inter_token_latency_ms_min": 12.969233647563104,
    "mean_inter_token_latency_ms_p25": 14.644854723818556,
    "mean_inter_token_latency_ms_p50": 14.70120383220281,
    "mean_inter_token_latency_ms_p75": 14.73719591618618,
    "mean_inter_token_latency_ms_p90": 14.795795856376312,
    "mean_inter_token_latency_ms_p95": 14.843224562355317,
    "mean_inter_token_latency_ms_p99": 16.110706731625314,
    "mean_inter_token_latency_ms_samples": 20,
    "mean_inter_token_latency_ms_stddev": 0.8045332337118766,
    "mean_time_per_output_token_max_ms": 16.110706731625314,
    "mean_time_per_output_token_mean_ci95_high_ms": 14.803233410672602,
    "mean_time_per_output_token_mean_ci95_low_ms": 14.098028938383509,
    "mean_time_per_output_token_mean_ms": 14.450631174528056,
    "mean_time_per_output_token_min_ms": 12.969233647563104,
    "mean_time_per_output_token_p25_ms": 14.644854723818556,
    "mean_time_per_output_token_p50_ms": 14.70120383220281,
    "mean_time_per_output_token_p75_ms": 14.73719591618618,
    "mean_time_per_output_token_p90_ms": 14.795795856376312,
    "mean_time_per_output_token_p95_ms": 14.843224562355317,
    "mean_time_per_output_token_p99_ms": 16.110706731625314,
    "mean_time_per_output_token_samples": 20,
    "mean_time_per_output_token_stddev_ms": 0.8045332337118766,
    "output_token_sources": {
      "usage": 20
    },
    "output_tokens": 5363,
    "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": 4200,
    "request_canary_passed": 0,
    "request_canary_samples": 0,
    "throughput_performance_eligible": true,
    "throughput_tok_s": 52.03134043413557,
    "time_to_first_output_max_ms": 1353.9961009519175,
    "time_to_first_output_mean_ci95_high_ms": 1325.4814744523924,
    "time_to_first_output_mean_ci95_low_ms": 1295.6871839643259,
    "time_to_first_output_mean_ms": 1310.5843292083591,
    "time_to_first_output_min_ms": 1248.8259819801897,
    "time_to_first_output_p25_ms": 1269.9585210066289,
    "time_to_first_output_p50_ms": 1317.3396319616586,
    "time_to_first_output_p75_ms": 1337.3597939498723,
    "time_to_first_output_p90_ms": 1343.2357760611922,
    "time_to_first_output_p95_ms": 1349.994505988434,
    "time_to_first_output_p99_ms": 1353.9961009519175,
    "time_to_first_output_samples": 20,
    "time_to_first_output_stddev_ms": 33.990846363618324,
    "tpot_max_ms": 16.110706731625314,
    "tpot_mean_ci95_high_ms": 14.803233410672602,
    "tpot_mean_ci95_low_ms": 14.098028938383509,
    "tpot_mean_ms": 14.450631174528056,
    "tpot_min_ms": 12.969233647563104,
    "tpot_p25_ms": 14.644854723818556,
    "tpot_p50_ms": 14.70120383220281,
    "tpot_p75_ms": 14.73719591618618,
    "tpot_p90_ms": 14.795795856376312,
    "tpot_p95_ms": 14.843224562355317,
    "tpot_p99_ms": 16.110706731625314,
    "tpot_samples": 20,
    "tpot_stddev_ms": 0.8045332337118766,
    "ttft_max_ms": 15555.633851909079,
    "ttft_mean_ci95_high_ms": 5739.623602973575,
    "ttft_mean_ci95_low_ms": 3144.3857250103747,
    "ttft_mean_ms": 4442.004663991975,
    "ttft_min_ms": 1522.680227062665,
    "ttft_p25_ms": 2528.0159800313413,
    "ttft_p50_ms": 3569.949577911757,
    "ttft_p75_ms": 5295.251138973981,
    "ttft_p90_ms": 6008.034082013182,
    "ttft_p95_ms": 6163.724130019546,
    "ttft_p99_ms": 15555.633851909079,
    "ttft_samples": 20,
    "ttft_stddev_ms": 2960.779751483684
  },
  "model": "llm.primary",
  "performance_eligible": true,
  "prompt_cache_mode": "unique",
  "request_canaries": false,
  "request_timings": [
    {
      "content_chunks": 40,
      "controlled_output": null,
      "decode_tok_s": 77.10555821375755,
      "e2e_ms": 3770.2961129834875,
      "effective_prefill_tok_s": 4787.5924328241745,
      "generation_ms": 2503.062093979679,
      "mean_inter_token_latency_ms": 12.969233647563104,
      "mean_time_per_output_token_ms": 12.969233647563104,
      "output_token_source": "usage",
      "output_tokens": 194,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6067,
      "reasoning_chunks": 151,
      "request_canary": null,
      "request_index": 0,
      "time_to_first_output_ms": 1267.2340190038085,
      "tpot_ms": 12.969233647563104,
      "ttft_ms": 3244.7394659975544,
      "visible_content_truncated": false,
      "visible_generation_ms": 525.5566469859332
    },
    {
      "content_chunks": 45,
      "controlled_output": null,
      "decode_tok_s": 76.74861602747636,
      "e2e_ms": 4818.922932026908,
      "effective_prefill_tok_s": 4857.362104511552,
      "generation_ms": 3570.096950046718,
      "mean_inter_token_latency_ms": 13.029550912579262,
      "mean_time_per_output_token_ms": 13.029550912579262,
      "output_token_source": "usage",
      "output_tokens": 275,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6066,
      "reasoning_chunks": 222,
      "request_canary": null,
      "request_index": 1,
      "time_to_first_output_ms": 1248.8259819801897,
      "tpot_ms": 13.029550912579262,
      "ttft_ms": 4214.308239985257,
      "visible_content_truncated": false,
      "visible_generation_ms": 604.6146920416504
    },
    {
      "content_chunks": 50,
      "controlled_output": null,
      "decode_tok_s": 76.82391773448269,
      "e2e_ms": 3992.5108778988943,
      "effective_prefill_tok_s": 4818.952977169809,
      "generation_ms": 2733.523701899685,
      "mean_inter_token_latency_ms": 13.016779532855642,
      "mean_time_per_output_token_ms": 13.016779532855642,
      "output_token_source": "usage",
      "output_tokens": 211,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6067,
      "reasoning_chunks": 153,
      "request_canary": null,
      "request_index": 2,
      "time_to_first_output_ms": 1258.9871759992093,
      "tpot_ms": 13.016779532855642,
      "ttft_ms": 3330.49860491883,
      "visible_content_truncated": false,
      "visible_generation_ms": 662.0122729800642
    },
    {
      "content_chunks": 54,
      "controlled_output": null,
      "decode_tok_s": 76.91526678458256,
      "e2e_ms": 6725.745756994002,
      "effective_prefill_tok_s": 4845.909899854215,
      "generation_ms": 5473.555739969015,
      "mean_inter_token_latency_ms": 13.00132004743234,
      "mean_time_per_output_token_ms": 13.00132004743234,
      "output_token_source": "usage",
      "output_tokens": 422,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6068,
      "reasoning_chunks": 347,
      "request_canary": null,
      "request_index": 3,
      "time_to_first_output_ms": 1252.1900170249864,
      "tpot_ms": 13.00132004743234,
      "ttft_ms": 6008.034082013182,
      "visible_content_truncated": false,
      "visible_generation_ms": 717.7116749808192
    },
    {
      "content_chunks": 46,
      "controlled_output": null,
      "decode_tok_s": 62.07052345115315,
      "e2e_ms": 3009.914848022163,
      "effective_prefill_tok_s": 4778.896239216872,
      "generation_ms": 1739.956327015534,
      "mean_inter_token_latency_ms": 16.110706731625314,
      "mean_time_per_output_token_ms": 16.110706731625314,
      "output_token_source": "usage",
      "output_tokens": 109,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6069,
      "reasoning_chunks": 60,
      "request_canary": null,
      "request_index": 4,
      "time_to_first_output_ms": 1269.9585210066289,
      "tpot_ms": 16.110706731625314,
      "ttft_ms": 2310.6059699784964,
      "visible_content_truncated": false,
      "visible_generation_ms": 699.3088780436665
    },
    {
      "content_chunks": 36,
      "controlled_output": null,
      "decode_tok_s": 67.85551374136774,
      "e2e_ms": 4120.3440389363095,
      "effective_prefill_tok_s": 4648.652489692725,
      "generation_ms": 2814.8044199915603,
      "mean_inter_token_latency_ms": 14.73719591618618,
      "mean_time_per_output_token_ms": 14.73719591618618,
      "output_token_source": "usage",
      "output_tokens": 192,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6069,
      "reasoning_chunks": 148,
      "request_canary": null,
      "request_index": 5,
      "time_to_first_output_ms": 1305.5396189447492,
      "tpot_ms": 14.73719591618618,
      "ttft_ms": 3569.949577911757,
      "visible_content_truncated": false,
      "visible_generation_ms": 550.3944610245526
    },
    {
      "content_chunks": 57,
      "controlled_output": null,
      "decode_tok_s": 68.02164036454703,
      "e2e_ms": 5684.09616407007,
      "effective_prefill_tok_s": 4658.756549512215,
      "generation_ms": 4380.958741996437,
      "mean_inter_token_latency_ms": 14.70120383220281,
      "mean_time_per_output_token_ms": 14.70120383220281,
      "output_token_source": "usage",
      "output_tokens": 299,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6071,
      "reasoning_chunks": 234,
      "request_canary": null,
      "request_index": 6,
      "time_to_first_output_ms": 1303.1374220736325,
      "tpot_ms": 14.70120383220281,
      "ttft_ms": 4831.583169056103,
      "visible_content_truncated": false,
      "visible_generation_ms": 852.5129950139672
    },
    {
      "content_chunks": 49,
      "controlled_output": null,
      "decode_tok_s": 67.87832819459257,
      "e2e_ms": 4576.066909939982,
      "effective_prefill_tok_s": 4650.2933196424565,
      "generation_ms": 3270.557862939313,
      "mean_inter_token_latency_ms": 14.732242625852761,
      "mean_time_per_output_token_ms": 14.732242625852761,
      "output_token_source": "usage",
      "output_tokens": 223,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6071,
      "reasoning_chunks": 163,
      "request_canary": null,
      "request_index": 7,
      "time_to_first_output_ms": 1305.509047000669,
      "tpot_ms": 14.732242625852761,
      "ttft_ms": 3839.918189914897,
      "visible_content_truncated": false,
      "visible_generation_ms": 736.1487200250849
    },
    {
      "content_chunks": 48,
      "controlled_output": null,
      "decode_tok_s": 68.28336769866272,
      "e2e_ms": 6238.2403099909425,
      "effective_prefill_tok_s": 4659.518706348461,
      "generation_ms": 4935.316041926853,
      "mean_inter_token_latency_ms": 14.644854723818556,
      "mean_time_per_output_token_ms": 14.644854723818556,
      "output_token_source": "usage",
      "output_tokens": 338,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6071,
      "reasoning_chunks": 282,
      "request_canary": null,
      "request_index": 8,
      "time_to_first_output_ms": 1302.9242680640891,
      "tpot_ms": 14.644854723818556,
      "ttft_ms": 5526.5970539767295,
      "visible_content_truncated": false,
      "visible_generation_ms": 711.643256014213
    },
    {
      "content_chunks": 51,
      "controlled_output": null,
      "decode_tok_s": 68.08941048313228,
      "e2e_ms": 6061.102248029783,
      "effective_prefill_tok_s": 4609.289702275295,
      "generation_ms": 4743.762616068125,
      "mean_inter_token_latency_ms": 14.68657156677438,
      "mean_time_per_output_token_ms": 14.68657156677438,
      "output_token_source": "usage",
      "output_tokens": 324,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6072,
      "reasoning_chunks": 256,
      "request_canary": null,
      "request_index": 9,
      "time_to_first_output_ms": 1317.3396319616586,
      "tpot_ms": 14.68657156677438,
      "ttft_ms": 5295.251138973981,
      "visible_content_truncated": false,
      "visible_generation_ms": 765.8511090558022
    },
    {
      "content_chunks": 40,
      "controlled_output": null,
      "decode_tok_s": 68.07571155037766,
      "e2e_ms": 5538.079696008936,
      "effective_prefill_tok_s": 4592.396594816607,
      "generation_ms": 4215.894236927852,
      "mean_inter_token_latency_ms": 14.689526957936767,
      "mean_time_per_output_token_ms": 14.689526957936767,
      "output_token_source": "usage",
      "output_tokens": 288,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6072,
      "reasoning_chunks": 241,
      "request_canary": null,
      "request_index": 10,
      "time_to_first_output_ms": 1322.1854590810835,
      "tpot_ms": 14.689526957936767,
      "ttft_ms": 4938.58081009239,
      "visible_content_truncated": false,
      "visible_generation_ms": 599.498885916546
    },
    {
      "content_chunks": 47,
      "controlled_output": null,
      "decode_tok_s": 68.00951177144326,
      "e2e_ms": 6874.3226629449055,
      "effective_prefill_tok_s": 4562.801932043573,
      "generation_ms": 5543.34224993363,
      "mean_inter_token_latency_ms": 14.703825596640929,
      "mean_time_per_output_token_ms": 14.703825596640929,
      "output_token_source": "usage",
      "output_tokens": 378,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6073,
      "reasoning_chunks": 317,
      "request_canary": null,
      "request_index": 11,
      "time_to_first_output_ms": 1330.9804130112752,
      "tpot_ms": 14.703825596640929,
      "ttft_ms": 6163.724130019546,
      "visible_content_truncated": false,
      "visible_generation_ms": 710.5985329253599
    },
    {
      "content_chunks": 47,
      "controlled_output": null,
      "decode_tok_s": 68.03405604046921,
      "e2e_ms": 6181.349255028181,
      "effective_prefill_tok_s": 4564.795301557347,
      "generation_ms": 4850.511923083104,
      "mean_inter_token_latency_ms": 14.69852097903971,
      "mean_time_per_output_token_ms": 14.69852097903971,
      "output_token_source": "usage",
      "output_tokens": 331,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6075,
      "reasoning_chunks": 270,
      "request_canary": null,
      "request_index": 12,
      "time_to_first_output_ms": 1330.8373319450766,
      "tpot_ms": 14.69852097903971,
      "ttft_ms": 5478.717378922738,
      "visible_content_truncated": false,
      "visible_generation_ms": 702.6318761054426
    },
    {
      "content_chunks": 54,
      "controlled_output": null,
      "decode_tok_s": 67.67080836016844,
      "e2e_ms": 3511.499513988383,
      "effective_prefill_tok_s": 4487.457530880858,
      "generation_ms": 2157.5034130364656,
      "mean_inter_token_latency_ms": 14.777420637236066,
      "mean_time_per_output_token_ms": 14.777420637236066,
      "output_token_source": "usage",
      "output_tokens": 147,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6076,
      "reasoning_chunks": 89,
      "request_canary": null,
      "request_index": 13,
      "time_to_first_output_ms": 1353.9961009519175,
      "tpot_ms": 14.777420637236066,
      "ttft_ms": 2703.752054949291,
      "visible_content_truncated": false,
      "visible_generation_ms": 807.7474590390921
    },
    {
      "content_chunks": 45,
      "controlled_output": null,
      "decode_tok_s": 68.15175800852141,
      "e2e_ms": 16230.59262696188,
      "effective_prefill_tok_s": 4543.279996518083,
      "generation_ms": 14893.232833012007,
      "mean_inter_token_latency_ms": 14.673135796070943,
      "mean_time_per_output_token_ms": 14.673135796070943,
      "output_token_source": "usage",
      "output_tokens": 1016,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6076,
      "reasoning_chunks": 926,
      "request_canary": null,
      "request_index": 14,
      "time_to_first_output_ms": 1337.3597939498723,
      "tpot_ms": 14.673135796070943,
      "ttft_ms": 15555.633851909079,
      "visible_content_truncated": false,
      "visible_generation_ms": 674.9587750528008
    },
    {
      "content_chunks": 43,
      "controlled_output": null,
      "decode_tok_s": 67.58676651848003,
      "e2e_ms": 2166.525493026711,
      "effective_prefill_tok_s": 4541.986156795802,
      "generation_ms": 828.5645679570735,
      "mean_inter_token_latency_ms": 14.795795856376312,
      "mean_time_per_output_token_ms": 14.795795856376312,
      "output_token_source": "usage",
      "output_tokens": 57,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6077,
      "reasoning_chunks": 12,
      "request_canary": null,
      "request_index": 15,
      "time_to_first_output_ms": 1337.9609250696376,
      "tpot_ms": 14.795795856376312,
      "ttft_ms": 1522.680227062665,
      "visible_content_truncated": false,
      "visible_generation_ms": 643.8452659640461
    },
    {
      "content_chunks": 63,
      "controlled_output": null,
      "decode_tok_s": 67.37080583798164,
      "e2e_ms": 3480.660113040358,
      "effective_prefill_tok_s": 4524.894369492368,
      "generation_ms": 2137.4243369791657,
      "mean_inter_token_latency_ms": 14.843224562355317,
      "mean_time_per_output_token_ms": 14.843224562355317,
      "output_token_source": "usage",
      "output_tokens": 145,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6078,
      "reasoning_chunks": 80,
      "request_canary": null,
      "request_index": 16,
      "time_to_first_output_ms": 1343.2357760611922,
      "tpot_ms": 14.843224562355317,
      "ttft_ms": 2528.0159800313413,
      "visible_content_truncated": false,
      "visible_generation_ms": 952.6441330090165
    },
    {
      "content_chunks": 51,
      "controlled_output": null,
      "decode_tok_s": 67.7430379109125,
      "e2e_ms": 3018.0625980719924,
      "effective_prefill_tok_s": 4502.981288467615,
      "generation_ms": 1668.0680920835584,
      "mean_inter_token_latency_ms": 14.761664531712906,
      "mean_time_per_output_token_ms": 14.761664531712906,
      "output_token_source": "usage",
      "output_tokens": 114,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6079,
      "reasoning_chunks": 60,
      "request_canary": null,
      "request_index": 17,
      "time_to_first_output_ms": 1349.994505988434,
      "tpot_ms": 14.761664531712906,
      "ttft_ms": 2253.957837005146,
      "visible_content_truncated": false,
      "visible_generation_ms": 764.1047610668465
    },
    {
      "content_chunks": 58,
      "controlled_output": null,
      "decode_tok_s": 67.9946278198847,
      "e2e_ms": 3289.2175640445203,
      "effective_prefill_tok_s": 4559.021965815532,
      "generation_ms": 1956.0368850361556,
      "mean_inter_token_latency_ms": 14.707044248392148,
      "mean_time_per_output_token_ms": 14.707044248392148,
      "output_token_source": "usage",
      "output_tokens": 134,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6078,
      "reasoning_chunks": 73,
      "request_canary": null,
      "request_index": 18,
      "time_to_first_output_ms": 1333.1806790083647,
      "tpot_ms": 14.707044248392148,
      "ttft_ms": 2427.651414065622,
      "visible_content_truncated": false,
      "visible_generation_ms": 861.5661499788985
    },
    {
      "content_chunks": 45,
      "controlled_output": null,
      "decode_tok_s": 67.87573815005274,
      "e2e_ms": 3771.222686045803,
      "effective_prefill_tok_s": 4534.772158255705,
      "generation_ms": 2430.9127900050953,
      "mean_inter_token_latency_ms": 14.732804787909668,
      "mean_time_per_output_token_ms": 14.732804787909668,
      "output_token_source": "usage",
      "output_tokens": 166,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6078,
      "reasoning_chunks": 116,
      "request_canary": null,
      "request_index": 19,
      "time_to_first_output_ms": 1340.3098960407078,
      "tpot_ms": 14.732804787909668,
      "ttft_ms": 3095.8941030548885,
      "visible_content_truncated": false,
      "visible_generation_ms": 675.3285829909146
    }
  ],
  "requests": 20,
  "response_words": 0,
  "responses_received": 20,
  "run_id": "benchmark-20261017T001533Z",
  "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": "2026-10-17T00:15:33Z",
  "started_at_unix_ns": 1792196133872384146,
  "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": 103072.49352510553
}