{
  "base_url": "redacted://routed-capability",
  "clock_domain_id": null,
  "completed": 19,
  "concurrency": 1,
  "configuration_fingerprint": null,
  "context_distribution": {
    "8192": 20
  },
  "context_seed": 941211,
  "context_tokens": 8192,
  "controlled_output_policy": "observe",
  "engine": "vllm-exl3-v84",
  "failed": 1,
  "failures": [
    {
      "elapsed_ms": 213753.30476195086,
      "error_message": "stream completed without visible content",
      "error_message_truncated": false,
      "error_type": "ValueError",
      "failure_class": "stream_validation_failed",
      "performance_eligible": false,
      "planned_context_tokens": 8192,
      "request_index": 16,
      "response_metadata": {
        "content_chunks": 0,
        "finish_reasons": [
          "length"
        ],
        "output_token_source": "usage",
        "reasoning_chunks": 15397,
        "stream_terminal_observed": true,
        "usage": {
          "completion_tokens": 16384,
          "prompt_tokens": 6078,
          "total_tokens": 22462
        },
        "visible_content_truncated": false
      },
      "response_received": true,
      "response_timing": null
    }
  ],
  "finished_at": "2026-10-17T00:44:20Z",
  "finished_at_unix_ns": 1792197860400103804,
  "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": false,
    "attempted_requests": 20,
    "excluded_requests": 1,
    "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": 19,
      "e2e": 19,
      "effective_prefill_tok_s": 19,
      "tpot": 19,
      "ttft": 19
    },
    "performance_eligible_requests": 19,
    "successful_requests": 19
  },
  "metrics": {
    "completion_tokens_max": 7143.0,
    "completion_tokens_mean": 606.3157894736842,
    "completion_tokens_mean_ci95_high": 1322.9660172500157,
    "completion_tokens_mean_ci95_low": -110.33443830264753,
    "completion_tokens_min": 113.0,
    "completion_tokens_p25": 129.0,
    "completion_tokens_p50": 190.0,
    "completion_tokens_p75": 327.0,
    "completion_tokens_p90": 943.0,
    "completion_tokens_p95": 7143.0,
    "completion_tokens_p99": 7143.0,
    "completion_tokens_samples": 19,
    "completion_tokens_stddev": 1593.7785309910387,
    "content_chunks": 1021,
    "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": 77.1477891036783,
    "decode_tok_s_mean": 76.83209955005228,
    "decode_tok_s_mean_ci95_high": 76.93134182672159,
    "decode_tok_s_mean_ci95_low": 76.73285727338296,
    "decode_tok_s_min": 76.24756299883587,
    "decode_tok_s_p25": 76.69658548819645,
    "decode_tok_s_p50": 76.84855418982426,
    "decode_tok_s_p75": 77.00947983266855,
    "decode_tok_s_p90": 77.14369611299222,
    "decode_tok_s_p95": 77.1477891036783,
    "decode_tok_s_p99": 77.1477891036783,
    "decode_tok_s_samples": 19,
    "decode_tok_s_stddev": 0.22070768108593802,
    "e2e_max_ms": 93908.35167490877,
    "e2e_mean_ci95_high_ms": 18444.94574130784,
    "e2e_mean_ci95_low_ms": -139.3805684769486,
    "e2e_mean_ms": 9152.782586415446,
    "e2e_min_ms": 2767.920478945598,
    "e2e_p25_ms": 2971.6930349823087,
    "e2e_p50_ms": 3737.337775994092,
    "e2e_p75_ms": 5503.6148900398985,
    "e2e_p90_ms": 13527.117270044982,
    "e2e_p95_ms": 93908.35167490877,
    "e2e_p99_ms": 93908.35167490877,
    "e2e_samples": 19,
    "e2e_stddev_ms": 20665.102121973487,
    "effective_prefill_tok_s_max": 4766.435371083118,
    "effective_prefill_tok_s_mean": 4686.159656745593,
    "effective_prefill_tok_s_mean_ci95_high": 4710.155025833501,
    "effective_prefill_tok_s_mean_ci95_low": 4662.164287657684,
    "effective_prefill_tok_s_min": 4573.465572506368,
    "effective_prefill_tok_s_p25": 4646.73129283785,
    "effective_prefill_tok_s_p50": 4686.298825090881,
    "effective_prefill_tok_s_p75": 4736.37670459169,
    "effective_prefill_tok_s_p90": 4752.369316164511,
    "effective_prefill_tok_s_p95": 4766.435371083118,
    "effective_prefill_tok_s_p99": 4766.435371083118,
    "effective_prefill_tok_s_samples": 19,
    "effective_prefill_tok_s_stddev": 53.36397396283203,
    "generation_max_ms": 92580.47461893875,
    "generation_mean_ci95_high_ms": 17145.627026282164,
    "generation_mean_ci95_low_ms": -1432.0060632414225,
    "generation_mean_ms": 7856.81048152037,
    "generation_min_ms": 1468.8994060270488,
    "generation_p25_ms": 1665.7357650110498,
    "generation_p50_ms": 2460.290560964495,
    "generation_p75_ms": 4229.916722048074,
    "generation_p90_ms": 12230.571696069092,
    "generation_p95_ms": 92580.47461893875,
    "generation_p99_ms": 92580.47461893875,
    "generation_samples": 19,
    "generation_stddev_ms": 20657.659501890466,
    "mean_inter_token_latency_ms_max": 13.11517326809865,
    "mean_inter_token_latency_ms_mean": 13.015495362505774,
    "mean_inter_token_latency_ms_mean_ci95_high": 13.032345634031117,
    "mean_inter_token_latency_ms_mean_ci95_low": 12.998645090980432,
    "mean_inter_token_latency_ms_min": 12.96213425709592,
    "mean_inter_token_latency_ms_p25": 12.985414291498504,
    "mean_inter_token_latency_ms_p50": 13.012606555093953,
    "mean_inter_token_latency_ms_p75": 13.038390087833823,
    "mean_inter_token_latency_ms_p90": 13.06496361630956,
    "mean_inter_token_latency_ms_p95": 13.11517326809865,
    "mean_inter_token_latency_ms_p99": 13.11517326809865,
    "mean_inter_token_latency_ms_samples": 19,
    "mean_inter_token_latency_ms_stddev": 0.03747379119907729,
    "mean_time_per_output_token_max_ms": 13.11517326809865,
    "mean_time_per_output_token_mean_ci95_high_ms": 13.032345634031117,
    "mean_time_per_output_token_mean_ci95_low_ms": 12.998645090980432,
    "mean_time_per_output_token_mean_ms": 13.015495362505774,
    "mean_time_per_output_token_min_ms": 12.96213425709592,
    "mean_time_per_output_token_p25_ms": 12.985414291498504,
    "mean_time_per_output_token_p50_ms": 13.012606555093953,
    "mean_time_per_output_token_p75_ms": 13.038390087833823,
    "mean_time_per_output_token_p90_ms": 13.06496361630956,
    "mean_time_per_output_token_p95_ms": 13.11517326809865,
    "mean_time_per_output_token_p99_ms": 13.11517326809865,
    "mean_time_per_output_token_samples": 19,
    "mean_time_per_output_token_stddev_ms": 0.03747379119907729,
    "output_token_sources": {
      "usage": 19
    },
    "output_tokens": 11520,
    "prefix_cache_hit_avg": null,
    "prompt_token_samples": 19,
    "prompt_tokens": 115375,
    "prompt_tokens_max": 6079.0,
    "prompt_tokens_mean": 6072.368421052632,
    "prompt_tokens_mean_ci95_high": 6074.229715260445,
    "prompt_tokens_mean_ci95_low": 6070.507126844818,
    "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": 6079.0,
    "prompt_tokens_p99": 6079.0,
    "prompt_tokens_samples": 19,
    "prompt_tokens_stddev": 4.139384365334609,
    "reasoning_chunks": 9619,
    "request_canary_passed": 0,
    "request_canary_samples": 0,
    "throughput_performance_eligible": false,
    "throughput_tok_s": null,
    "time_to_first_output_max_ms": 1327.8770559700206,
    "time_to_first_output_mean_ci95_high_ms": 1302.7750050838463,
    "time_to_first_output_mean_ci95_low_ms": 1289.1692047063052,
    "time_to_first_output_mean_ms": 1295.9721048950757,
    "time_to_first_output_min_ms": 1273.6981679918244,
    "time_to_first_output_p25_ms": 1281.9926240481436,
    "time_to_first_output_p50_ms": 1296.5455739758909,
    "time_to_first_output_p75_ms": 1306.0793959302828,
    "time_to_first_output_p90_ms": 1315.639007021673,
    "time_to_first_output_p95_ms": 1327.8770559700206,
    "time_to_first_output_p99_ms": 1327.8770559700206,
    "time_to_first_output_samples": 19,
    "time_to_first_output_stddev_ms": 15.129160431553228,
    "tpot_max_ms": 13.11517326809865,
    "tpot_mean_ci95_high_ms": 13.032345634031117,
    "tpot_mean_ci95_low_ms": 12.998645090980432,
    "tpot_mean_ms": 13.015495362505774,
    "tpot_min_ms": 12.96213425709592,
    "tpot_p25_ms": 12.985414291498504,
    "tpot_p50_ms": 13.012606555093953,
    "tpot_p75_ms": 13.038390087833823,
    "tpot_p90_ms": 13.06496361630956,
    "tpot_p95_ms": 13.11517326809865,
    "tpot_p99_ms": 13.11517326809865,
    "tpot_samples": 19,
    "tpot_stddev_ms": 0.03747379119907729,
    "ttft_max_ms": 92902.18022093177,
    "ttft_mean_ci95_high_ms": 17695.958425098448,
    "ttft_mean_ci95_low_ms": -816.1578349878982,
    "ttft_mean_ms": 8439.900295055275,
    "ttft_min_ms": 1993.4106380678713,
    "ttft_p25_ms": 2327.7286669472232,
    "ttft_p50_ms": 3084.16227507405,
    "ttft_p75_ms": 4809.6815820317715,
    "ttft_p90_ms": 12441.26585801132,
    "ttft_p95_ms": 92902.18022093177,
    "ttft_p99_ms": 92902.18022093177,
    "ttft_samples": 19,
    "ttft_stddev_ms": 20584.80714510013
  },
  "model": "llm.primary",
  "performance_eligible": false,
  "prompt_cache_mode": "unique",
  "request_canaries": false,
  "request_timings": [
    {
      "content_chunks": 53,
      "controlled_output": null,
      "decode_tok_s": 77.00947983266855,
      "e2e_ms": 5734.174930024892,
      "effective_prefill_tok_s": 4739.185343714719,
      "generation_ms": 4453.997101983987,
      "mean_inter_token_latency_ms": 12.985414291498504,
      "mean_time_per_output_token_ms": 12.985414291498504,
      "output_token_source": "usage",
      "output_tokens": 344,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6067,
      "reasoning_chunks": 284,
      "request_canary": null,
      "request_index": 0,
      "time_to_first_output_ms": 1280.1778280409053,
      "tpot_ms": 12.985414291498504,
      "ttft_ms": 5033.540802076459,
      "visible_content_truncated": false,
      "visible_generation_ms": 700.6341279484332
    },
    {
      "content_chunks": 45,
      "controlled_output": null,
      "decode_tok_s": 76.85690471763971,
      "e2e_ms": 3444.0541529329494,
      "effective_prefill_tok_s": 4629.761975990769,
      "generation_ms": 2133.8356079068035,
      "mean_inter_token_latency_ms": 13.011192731139046,
      "mean_time_per_output_token_ms": 13.011192731139046,
      "output_token_source": "usage",
      "output_tokens": 165,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6066,
      "reasoning_chunks": 114,
      "request_canary": null,
      "request_index": 1,
      "time_to_first_output_ms": 1310.2185450261459,
      "tpot_ms": 13.011192731139046,
      "ttft_ms": 2847.5867320084944,
      "visible_content_truncated": false,
      "visible_generation_ms": 596.4674209244549
    },
    {
      "content_chunks": 48,
      "controlled_output": null,
      "decode_tok_s": 76.69658548819645,
      "e2e_ms": 4255.219927988946,
      "effective_prefill_tok_s": 4730.726059184784,
      "generation_ms": 2972.752940026112,
      "mean_inter_token_latency_ms": 13.038390087833823,
      "mean_time_per_output_token_ms": 13.038390087833823,
      "output_token_source": "usage",
      "output_tokens": 229,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6067,
      "reasoning_chunks": 177,
      "request_canary": null,
      "request_index": 2,
      "time_to_first_output_ms": 1282.4669879628345,
      "tpot_ms": 13.038390087833823,
      "ttft_ms": 3610.236744978465,
      "visible_content_truncated": false,
      "visible_generation_ms": 644.9831830104813
    },
    {
      "content_chunks": 47,
      "controlled_output": null,
      "decode_tok_s": 76.80955721742528,
      "e2e_ms": 5600.255277007818,
      "effective_prefill_tok_s": 4700.612093418905,
      "generation_ms": 4309.359563980252,
      "mean_inter_token_latency_ms": 13.019213184230367,
      "mean_time_per_output_token_ms": 13.019213184230367,
      "output_token_source": "usage",
      "output_tokens": 332,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6068,
      "reasoning_chunks": 270,
      "request_canary": null,
      "request_index": 3,
      "time_to_first_output_ms": 1290.8957130275667,
      "tpot_ms": 13.019213184230367,
      "ttft_ms": 4967.274060938507,
      "visible_content_truncated": false,
      "visible_generation_ms": 632.9812160693109
    },
    {
      "content_chunks": 49,
      "controlled_output": null,
      "decode_tok_s": 76.84855418982426,
      "e2e_ms": 2971.6930349823087,
      "effective_prefill_tok_s": 4646.73129283785,
      "generation_ms": 1665.613639052026,
      "mean_inter_token_latency_ms": 13.012606555093953,
      "mean_time_per_output_token_ms": 13.012606555093953,
      "output_token_source": "usage",
      "output_tokens": 129,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6069,
      "reasoning_chunks": 76,
      "request_canary": null,
      "request_index": 4,
      "time_to_first_output_ms": 1306.0793959302828,
      "tpot_ms": 13.012606555093953,
      "ttft_ms": 2327.7286669472232,
      "visible_content_truncated": false,
      "visible_generation_ms": 643.9643680350855
    },
    {
      "content_chunks": 49,
      "controlled_output": null,
      "decode_tok_s": 76.82019473582312,
      "e2e_ms": 3737.337775994092,
      "effective_prefill_tok_s": 4752.369316164511,
      "generation_ms": 2460.290560964495,
      "mean_inter_token_latency_ms": 13.017410375473519,
      "mean_time_per_output_token_ms": 13.017410375473519,
      "output_token_source": "usage",
      "output_tokens": 190,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6069,
      "reasoning_chunks": 130,
      "request_canary": null,
      "request_index": 5,
      "time_to_first_output_ms": 1277.0472150295973,
      "tpot_ms": 13.017410375473519,
      "ttft_ms": 3084.16227507405,
      "visible_content_truncated": false,
      "visible_generation_ms": 653.1755009200424
    },
    {
      "content_chunks": 49,
      "controlled_output": null,
      "decode_tok_s": 76.6689895130071,
      "e2e_ms": 3703.948606038466,
      "effective_prefill_tok_s": 4655.5987324948765,
      "generation_ms": 2399.92728701327,
      "mean_inter_token_latency_ms": 13.043083081593858,
      "mean_time_per_output_token_ms": 13.043083081593858,
      "output_token_source": "usage",
      "output_tokens": 185,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6071,
      "reasoning_chunks": 130,
      "request_canary": null,
      "request_index": 6,
      "time_to_first_output_ms": 1304.0213190251961,
      "tpot_ms": 13.043083081593858,
      "ttft_ms": 3048.106288071722,
      "visible_content_truncated": false,
      "visible_generation_ms": 655.8423179667443
    },
    {
      "content_chunks": 54,
      "controlled_output": null,
      "decode_tok_s": 77.1477891036783,
      "e2e_ms": 3043.756252969615,
      "effective_prefill_tok_s": 4739.614014155624,
      "generation_ms": 1762.8502589650452,
      "mean_inter_token_latency_ms": 12.96213425709592,
      "mean_time_per_output_token_ms": 12.96213425709592,
      "output_token_source": "usage",
      "output_tokens": 137,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6071,
      "reasoning_chunks": 78,
      "request_canary": null,
      "request_index": 7,
      "time_to_first_output_ms": 1280.90599400457,
      "tpot_ms": 12.96213425709592,
      "ttft_ms": 2336.8870719568804,
      "visible_content_truncated": false,
      "visible_generation_ms": 706.8691810127348
    },
    {
      "content_chunks": 53,
      "controlled_output": null,
      "decode_tok_s": 77.07007523357453,
      "e2e_ms": 5503.6148900398985,
      "effective_prefill_tok_s": 4766.435371083118,
      "generation_ms": 4229.916722048074,
      "mean_inter_token_latency_ms": 12.975204668859122,
      "mean_time_per_output_token_ms": 12.975204668859122,
      "output_token_source": "usage",
      "output_tokens": 327,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6071,
      "reasoning_chunks": 261,
      "request_canary": null,
      "request_index": 8,
      "time_to_first_output_ms": 1273.6981679918244,
      "tpot_ms": 12.975204668859122,
      "ttft_ms": 4809.6815820317715,
      "visible_content_truncated": false,
      "visible_generation_ms": 693.9333080081269
    },
    {
      "content_chunks": 56,
      "controlled_output": null,
      "decode_tok_s": 76.89482401761899,
      "e2e_ms": 4351.119876955636,
      "effective_prefill_tok_s": 4736.37670459169,
      "generation_ms": 3069.127252907492,
      "mean_inter_token_latency_ms": 13.004776495370729,
      "mean_time_per_output_token_ms": 13.004776495370729,
      "output_token_source": "usage",
      "output_tokens": 237,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6072,
      "reasoning_chunks": 173,
      "request_canary": null,
      "request_index": 9,
      "time_to_first_output_ms": 1281.9926240481436,
      "tpot_ms": 13.004776495370729,
      "ttft_ms": 3609.741537948139,
      "visible_content_truncated": false,
      "visible_generation_ms": 741.3783390074968
    },
    {
      "content_chunks": 52,
      "controlled_output": null,
      "decode_tok_s": 76.85022547333404,
      "e2e_ms": 4613.20264195092,
      "effective_prefill_tok_s": 4736.172830252052,
      "generation_ms": 3331.1548329656944,
      "mean_inter_token_latency_ms": 13.012323566272244,
      "mean_time_per_output_token_ms": 13.012323566272244,
      "output_token_source": "usage",
      "output_tokens": 257,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6072,
      "reasoning_chunks": 198,
      "request_canary": null,
      "request_index": 10,
      "time_to_first_output_ms": 1282.0478089852259,
      "tpot_ms": 13.012323566272244,
      "ttft_ms": 3921.9108229735866,
      "visible_content_truncated": false,
      "visible_generation_ms": 691.2918189773336
    },
    {
      "content_chunks": 76,
      "controlled_output": null,
      "decode_tok_s": 77.14369611299222,
      "e2e_ms": 93908.35167490877,
      "effective_prefill_tok_s": 4573.465572506368,
      "generation_ms": 92580.47461893875,
      "mean_inter_token_latency_ms": 12.96282198528966,
      "mean_time_per_output_token_ms": 12.96282198528966,
      "output_token_source": "usage",
      "output_tokens": 7143,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6073,
      "reasoning_chunks": 6358,
      "request_canary": null,
      "request_index": 11,
      "time_to_first_output_ms": 1327.8770559700206,
      "tpot_ms": 12.96282198528966,
      "ttft_ms": 92902.18022093177,
      "visible_content_truncated": false,
      "visible_generation_ms": 1006.1714539770037
    },
    {
      "content_chunks": 46,
      "controlled_output": null,
      "decode_tok_s": 76.54058819969897,
      "e2e_ms": 3221.137659973465,
      "effective_prefill_tok_s": 4624.50900620027,
      "generation_ms": 1907.4846879811957,
      "mean_inter_token_latency_ms": 13.06496361630956,
      "mean_time_per_output_token_ms": 13.06496361630956,
      "output_token_source": "usage",
      "output_tokens": 147,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6075,
      "reasoning_chunks": 97,
      "request_canary": null,
      "request_index": 12,
      "time_to_first_output_ms": 1313.6529719922692,
      "tpot_ms": 13.06496361630956,
      "ttft_ms": 2603.824606980197,
      "visible_content_truncated": false,
      "visible_generation_ms": 617.3130529932678
    },
    {
      "content_chunks": 49,
      "controlled_output": null,
      "decode_tok_s": 76.84291992082603,
      "e2e_ms": 2970.286839059554,
      "effective_prefill_tok_s": 4657.540912632823,
      "generation_ms": 1665.7357650110498,
      "mean_inter_token_latency_ms": 13.013560664148827,
      "mean_time_per_output_token_ms": 13.013560664148827,
      "output_token_source": "usage",
      "output_tokens": 129,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6076,
      "reasoning_chunks": 76,
      "request_canary": null,
      "request_index": 13,
      "time_to_first_output_ms": 1304.5510740485042,
      "tpot_ms": 13.013560664148827,
      "ttft_ms": 2322.2891950281337,
      "visible_content_truncated": false,
      "visible_generation_ms": 647.9976440314204
    },
    {
      "content_chunks": 82,
      "controlled_output": null,
      "decode_tok_s": 77.02011184830869,
      "e2e_ms": 13527.117270044982,
      "effective_prefill_tok_s": 4686.298825090881,
      "generation_ms": 12230.571696069092,
      "mean_inter_token_latency_ms": 12.98362175803513,
      "mean_time_per_output_token_ms": 12.98362175803513,
      "output_token_source": "usage",
      "output_tokens": 943,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6076,
      "reasoning_chunks": 804,
      "request_canary": null,
      "request_index": 14,
      "time_to_first_output_ms": 1296.5455739758909,
      "tpot_ms": 12.98362175803513,
      "ttft_ms": 12441.26585801132,
      "visible_content_truncated": false,
      "visible_generation_ms": 1085.8514120336622
    },
    {
      "content_chunks": 66,
      "controlled_output": null,
      "decode_tok_s": 76.60805942463391,
      "e2e_ms": 2867.0117030851543,
      "effective_prefill_tok_s": 4672.470080903284,
      "generation_ms": 1566.4148250361905,
      "mean_inter_token_latency_ms": 13.053456875301586,
      "mean_time_per_output_token_ms": 13.053456875301586,
      "output_token_source": "usage",
      "output_tokens": 121,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6077,
      "reasoning_chunks": 52,
      "request_canary": null,
      "request_index": 15,
      "time_to_first_output_ms": 1300.5968780489638,
      "tpot_ms": 13.053456875301586,
      "ttft_ms": 1993.4106380678713,
      "visible_content_truncated": false,
      "visible_generation_ms": 873.601065017283
    },
    {
      "content_chunks": 50,
      "controlled_output": null,
      "decode_tok_s": 76.98189918287115,
      "e2e_ms": 4758.006665040739,
      "effective_prefill_tok_s": 4620.568383542811,
      "generation_ms": 3442.367658019066,
      "mean_inter_token_latency_ms": 12.99006663403421,
      "mean_time_per_output_token_ms": 12.99006663403421,
      "output_token_source": "usage",
      "output_tokens": 266,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6079,
      "reasoning_chunks": 206,
      "request_canary": null,
      "request_index": 17,
      "time_to_first_output_ms": 1315.639007021673,
      "tpot_ms": 12.99006663403421,
      "ttft_ms": 4096.703800023533,
      "visible_content_truncated": false,
      "visible_generation_ms": 661.3028650172055
    },
    {
      "content_chunks": 51,
      "controlled_output": null,
      "decode_tok_s": 76.24756299883587,
      "e2e_ms": 2767.920478945598,
      "effective_prefill_tok_s": 4678.907930526775,
      "generation_ms": 1468.8994060270488,
      "mean_inter_token_latency_ms": 13.11517326809865,
      "mean_time_per_output_token_ms": 13.11517326809865,
      "output_token_source": "usage",
      "output_tokens": 113,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6078,
      "reasoning_chunks": 59,
      "request_canary": null,
      "request_index": 18,
      "time_to_first_output_ms": 1299.0210729185492,
      "tpot_ms": 13.11517326809865,
      "ttft_ms": 2084.78626399301,
      "visible_content_truncated": false,
      "visible_generation_ms": 683.1342149525881
    },
    {
      "content_chunks": 46,
      "controlled_output": null,
      "decode_tok_s": 76.7518742400362,
      "e2e_ms": 2924.659483949654,
      "effective_prefill_tok_s": 4689.689032874148,
      "generation_ms": 1628.6247239913791,
      "mean_inter_token_latency_ms": 13.028997791931033,
      "mean_time_per_output_token_ms": 13.028997791931033,
      "output_token_source": "usage",
      "output_tokens": 126,
      "planned_context_tokens": 8192,
      "prompt_tokens": 6078,
      "reasoning_chunks": 76,
      "request_canary": null,
      "request_index": 19,
      "time_to_first_output_ms": 1296.0347599582747,
      "tpot_ms": 13.028997791931033,
      "ttft_ms": 2316.788438009098,
      "visible_content_truncated": false,
      "visible_generation_ms": 607.8710459405556
    }
  ],
  "requests": 20,
  "response_words": 0,
  "responses_received": 20,
  "run_id": "benchmark-20261017T003752Z",
  "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": "2026-10-17T00:37:52Z",
  "started_at_unix_ns": 1792197472731242349,
  "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": 387668.85691403877
}