{
  "base_url": "http://127.0.0.1:8001/v1",
  "clock_domain_id": null,
  "completed": 1,
  "concurrency": 1,
  "configuration_fingerprint": null,
  "context_distribution": {
    "4096": 10
  },
  "context_seed": 29,
  "context_tokens": 4096,
  "controlled_output_policy": "observe",
  "engine": "sglang-glm53-sm120-v0.1.1-rc14",
  "failed": 9,
  "failures": [
    {
      "elapsed_ms": 858.9840960012225,
      "error_message": "request canary validation failed",
      "error_message_truncated": false,
      "error_type": "ValueError",
      "failure_class": "request_canary_failed",
      "performance_eligible": false,
      "planned_context_tokens": 4096,
      "request_canary": {
        "capture_complete": true,
        "foreign_marker_count": 0,
        "marker": "ANVIL_REQ_29_00000",
        "marker_at_start": false,
        "observed_marker_count": 1,
        "passed": false
      },
      "request_index": 0,
      "response_metadata": {
        "content_chunks": 18,
        "finish_reasons": [
          "stop"
        ],
        "output_token_source": "usage",
        "reasoning_chunks": 0,
        "stream_terminal_observed": true,
        "usage": {
          "completion_tokens": 44,
          "prompt_tokens": 2979,
          "total_tokens": 3023
        },
        "visible_content_truncated": false
      },
      "response_received": true,
      "response_timing": {
        "content_chunks": 18,
        "controlled_output": null,
        "decode_tok_s": 135.27319924770177,
        "e2e_ms": 858.567454997683,
        "effective_prefill_tok_s": 5509.603966578468,
        "generation_ms": 317.8752349995193,
        "mean_inter_token_latency_ms": 7.392447325570217,
        "mean_time_per_output_token_ms": 7.392447325570217,
        "output_token_source": "usage",
        "output_tokens": 44,
        "planned_context_tokens": 4096,
        "prompt_tokens": 2979,
        "reasoning_chunks": 0,
        "request_canary": {
          "capture_complete": true,
          "foreign_marker_count": 0,
          "marker": "ANVIL_REQ_29_00000",
          "marker_at_start": false,
          "observed_marker_count": 1,
          "passed": false
        },
        "request_index": 0,
        "time_to_first_output_ms": 540.6922199981636,
        "tpot_ms": 7.392447325570217,
        "ttft_ms": 540.6928999982483,
        "visible_content_truncated": false,
        "visible_generation_ms": 317.8745549994346
      }
    },
    {
      "elapsed_ms": 757.2265359995072,
      "error_message": "request canary validation failed",
      "error_message_truncated": false,
      "error_type": "ValueError",
      "failure_class": "request_canary_failed",
      "performance_eligible": false,
      "planned_context_tokens": 4096,
      "request_canary": {
        "capture_complete": true,
        "foreign_marker_count": 0,
        "marker": "ANVIL_REQ_29_00002",
        "marker_at_start": false,
        "observed_marker_count": 1,
        "passed": false
      },
      "request_index": 2,
      "response_metadata": {
        "content_chunks": 14,
        "finish_reasons": [
          "stop"
        ],
        "output_token_source": "usage",
        "reasoning_chunks": 0,
        "stream_terminal_observed": true,
        "usage": {
          "completion_tokens": 35,
          "prompt_tokens": 2979,
          "total_tokens": 3014
        },
        "visible_content_truncated": false
      },
      "response_received": true,
      "response_timing": {
        "content_chunks": 14,
        "controlled_output": null,
        "decode_tok_s": 162.7992749264488,
        "e2e_ms": 757.095704000676,
        "effective_prefill_tok_s": 5433.656841995607,
        "generation_ms": 208.84613899761462,
        "mean_inter_token_latency_ms": 6.142533499929842,
        "mean_time_per_output_token_ms": 6.142533499929842,
        "output_token_source": "usage",
        "output_tokens": 35,
        "planned_context_tokens": 4096,
        "prompt_tokens": 2979,
        "reasoning_chunks": 0,
        "request_canary": {
          "capture_complete": true,
          "foreign_marker_count": 0,
          "marker": "ANVIL_REQ_29_00002",
          "marker_at_start": false,
          "observed_marker_count": 1,
          "passed": false
        },
        "request_index": 2,
        "time_to_first_output_ms": 548.2495650030614,
        "tpot_ms": 6.142533499929842,
        "ttft_ms": 548.2498450001003,
        "visible_content_truncated": false,
        "visible_generation_ms": 208.84585900057573
      }
    },
    {
      "elapsed_ms": 765.0122609993559,
      "error_message": "request canary validation failed",
      "error_message_truncated": false,
      "error_type": "ValueError",
      "failure_class": "request_canary_failed",
      "performance_eligible": false,
      "planned_context_tokens": 4096,
      "request_canary": {
        "capture_complete": true,
        "foreign_marker_count": 0,
        "marker": "ANVIL_REQ_29_00003",
        "marker_at_start": false,
        "observed_marker_count": 1,
        "passed": false
      },
      "request_index": 3,
      "response_metadata": {
        "content_chunks": 15,
        "finish_reasons": [
          "stop"
        ],
        "output_token_source": "usage",
        "reasoning_chunks": 0,
        "stream_terminal_observed": true,
        "usage": {
          "completion_tokens": 36,
          "prompt_tokens": 2978,
          "total_tokens": 3014
        },
        "visible_content_truncated": false
      },
      "response_received": true,
      "response_timing": {
        "content_chunks": 15,
        "controlled_output": null,
        "decode_tok_s": 156.9836490671077,
        "e2e_ms": 764.8459109987016,
        "effective_prefill_tok_s": 5495.552344135607,
        "generation_ms": 222.9531560005853,
        "mean_inter_token_latency_ms": 6.370090171445294,
        "mean_time_per_output_token_ms": 6.370090171445294,
        "output_token_source": "usage",
        "output_tokens": 36,
        "planned_context_tokens": 4096,
        "prompt_tokens": 2978,
        "reasoning_chunks": 0,
        "request_canary": {
          "capture_complete": true,
          "foreign_marker_count": 0,
          "marker": "ANVIL_REQ_29_00003",
          "marker_at_start": false,
          "observed_marker_count": 1,
          "passed": false
        },
        "request_index": 3,
        "time_to_first_output_ms": 541.8927549981163,
        "tpot_ms": 6.370090171445294,
        "ttft_ms": 541.8930050000199,
        "visible_content_truncated": false,
        "visible_generation_ms": 222.95290599868167
      }
    },
    {
      "elapsed_ms": 773.5852199984947,
      "error_message": "request canary validation failed",
      "error_message_truncated": false,
      "error_type": "ValueError",
      "failure_class": "request_canary_failed",
      "performance_eligible": false,
      "planned_context_tokens": 4096,
      "request_canary": {
        "capture_complete": true,
        "foreign_marker_count": 0,
        "marker": "ANVIL_REQ_29_00004",
        "marker_at_start": false,
        "observed_marker_count": 1,
        "passed": false
      },
      "request_index": 4,
      "response_metadata": {
        "content_chunks": 14,
        "finish_reasons": [
          "stop"
        ],
        "output_token_source": "usage",
        "reasoning_chunks": 0,
        "stream_terminal_observed": true,
        "usage": {
          "completion_tokens": 36,
          "prompt_tokens": 2978,
          "total_tokens": 3014
        },
        "visible_content_truncated": false
      },
      "response_received": true,
      "response_timing": {
        "content_chunks": 14,
        "controlled_output": null,
        "decode_tok_s": 156.89052869454844,
        "e2e_ms": 773.4670000027108,
        "effective_prefill_tok_s": 5410.792204428936,
        "generation_ms": 223.08548700311803,
        "mean_inter_token_latency_ms": 6.373871057231944,
        "mean_time_per_output_token_ms": 6.373871057231944,
        "output_token_source": "usage",
        "output_tokens": 36,
        "planned_context_tokens": 4096,
        "prompt_tokens": 2978,
        "reasoning_chunks": 0,
        "request_canary": {
          "capture_complete": true,
          "foreign_marker_count": 0,
          "marker": "ANVIL_REQ_29_00004",
          "marker_at_start": false,
          "observed_marker_count": 1,
          "passed": false
        },
        "request_index": 4,
        "time_to_first_output_ms": 550.3815129995928,
        "tpot_ms": 6.373871057231944,
        "ttft_ms": 550.382093002554,
        "visible_content_truncated": false,
        "visible_generation_ms": 223.08490700015682
      }
    },
    {
      "elapsed_ms": 869.859672999155,
      "error_message": "request canary validation failed",
      "error_message_truncated": false,
      "error_type": "ValueError",
      "failure_class": "request_canary_failed",
      "performance_eligible": false,
      "planned_context_tokens": 4096,
      "request_canary": {
        "capture_complete": true,
        "foreign_marker_count": 0,
        "marker": "ANVIL_REQ_29_00005",
        "marker_at_start": false,
        "observed_marker_count": 1,
        "passed": false
      },
      "request_index": 5,
      "response_metadata": {
        "content_chunks": 20,
        "finish_reasons": [
          "stop"
        ],
        "output_token_source": "usage",
        "reasoning_chunks": 0,
        "stream_terminal_observed": true,
        "usage": {
          "completion_tokens": 49,
          "prompt_tokens": 2977,
          "total_tokens": 3026
        },
        "visible_content_truncated": false
      },
      "response_received": true,
      "response_timing": {
        "content_chunks": 20,
        "controlled_output": null,
        "decode_tok_s": 148.14970097730057,
        "e2e_ms": 869.7417319999659,
        "effective_prefill_tok_s": 5454.927304436845,
        "generation_ms": 323.99660399823915,
        "mean_inter_token_latency_ms": 6.749929249963316,
        "mean_time_per_output_token_ms": 6.749929249963316,
        "output_token_source": "usage",
        "output_tokens": 49,
        "planned_context_tokens": 4096,
        "prompt_tokens": 2977,
        "reasoning_chunks": 0,
        "request_canary": {
          "capture_complete": true,
          "foreign_marker_count": 0,
          "marker": "ANVIL_REQ_29_00005",
          "marker_at_start": false,
          "observed_marker_count": 1,
          "passed": false
        },
        "request_index": 5,
        "time_to_first_output_ms": 545.7451280017267,
        "tpot_ms": 6.749929249963316,
        "ttft_ms": 545.7453980015998,
        "visible_content_truncated": false,
        "visible_generation_ms": 323.996333998366
      }
    },
    {
      "elapsed_ms": 765.1389820020995,
      "error_message": "request canary validation failed",
      "error_message_truncated": false,
      "error_type": "ValueError",
      "failure_class": "request_canary_failed",
      "performance_eligible": false,
      "planned_context_tokens": 4096,
      "request_canary": {
        "capture_complete": true,
        "foreign_marker_count": 0,
        "marker": "ANVIL_REQ_29_00006",
        "marker_at_start": false,
        "observed_marker_count": 1,
        "passed": false
      },
      "request_index": 6,
      "response_metadata": {
        "content_chunks": 14,
        "finish_reasons": [
          "stop"
        ],
        "output_token_source": "usage",
        "reasoning_chunks": 0,
        "stream_terminal_observed": true,
        "usage": {
          "completion_tokens": 37,
          "prompt_tokens": 2975,
          "total_tokens": 3012
        },
        "visible_content_truncated": false
      },
      "response_received": true,
      "response_timing": {
        "content_chunks": 14,
        "controlled_output": null,
        "decode_tok_s": 161.77258462927801,
        "e2e_ms": 765.0337920022139,
        "effective_prefill_tok_s": 5483.879297163222,
        "generation_ms": 222.53461600121227,
        "mean_inter_token_latency_ms": 6.181517111144785,
        "mean_time_per_output_token_ms": 6.181517111144785,
        "output_token_source": "usage",
        "output_tokens": 37,
        "planned_context_tokens": 4096,
        "prompt_tokens": 2975,
        "reasoning_chunks": 0,
        "request_canary": {
          "capture_complete": true,
          "foreign_marker_count": 0,
          "marker": "ANVIL_REQ_29_00006",
          "marker_at_start": false,
          "observed_marker_count": 1,
          "passed": false
        },
        "request_index": 6,
        "time_to_first_output_ms": 542.4991760010016,
        "tpot_ms": 6.181517111144785,
        "ttft_ms": 542.4994060012978,
        "visible_content_truncated": false,
        "visible_generation_ms": 222.53438600091613
      }
    },
    {
      "elapsed_ms": 833.4672109995154,
      "error_message": "request canary validation failed",
      "error_message_truncated": false,
      "error_type": "ValueError",
      "failure_class": "request_canary_failed",
      "performance_eligible": false,
      "planned_context_tokens": 4096,
      "request_canary": {
        "capture_complete": true,
        "foreign_marker_count": 0,
        "marker": "ANVIL_REQ_29_00007",
        "marker_at_start": false,
        "observed_marker_count": 1,
        "passed": false
      },
      "request_index": 7,
      "response_metadata": {
        "content_chunks": 19,
        "finish_reasons": [
          "stop"
        ],
        "output_token_source": "usage",
        "reasoning_chunks": 0,
        "stream_terminal_observed": true,
        "usage": {
          "completion_tokens": 51,
          "prompt_tokens": 2974,
          "total_tokens": 3025
        },
        "visible_content_truncated": false
      },
      "response_received": true,
      "response_timing": {
        "content_chunks": 19,
        "controlled_output": null,
        "decode_tok_s": 172.12744307610717,
        "e2e_ms": 833.3555800018075,
        "effective_prefill_tok_s": 5478.259637136911,
        "generation_ms": 290.48244200021145,
        "mean_inter_token_latency_ms": 5.809648840004229,
        "mean_time_per_output_token_ms": 5.809648840004229,
        "output_token_source": "usage",
        "output_tokens": 51,
        "planned_context_tokens": 4096,
        "prompt_tokens": 2974,
        "reasoning_chunks": 0,
        "request_canary": {
          "capture_complete": true,
          "foreign_marker_count": 0,
          "marker": "ANVIL_REQ_29_00007",
          "marker_at_start": false,
          "observed_marker_count": 1,
          "passed": false
        },
        "request_index": 7,
        "time_to_first_output_ms": 542.8731380015961,
        "tpot_ms": 5.809648840004229,
        "ttft_ms": 542.873378002696,
        "visible_content_truncated": false,
        "visible_generation_ms": 290.48220199911157
      }
    },
    {
      "elapsed_ms": 872.0722590005607,
      "error_message": "request canary validation failed",
      "error_message_truncated": false,
      "error_type": "ValueError",
      "failure_class": "request_canary_failed",
      "performance_eligible": false,
      "planned_context_tokens": 4096,
      "request_canary": {
        "capture_complete": true,
        "foreign_marker_count": 0,
        "marker": "ANVIL_REQ_29_00008",
        "marker_at_start": false,
        "observed_marker_count": 1,
        "passed": false
      },
      "request_index": 8,
      "response_metadata": {
        "content_chunks": 21,
        "finish_reasons": [
          "stop"
        ],
        "output_token_source": "usage",
        "reasoning_chunks": 0,
        "stream_terminal_observed": true,
        "usage": {
          "completion_tokens": 52,
          "prompt_tokens": 2974,
          "total_tokens": 3026
        },
        "visible_content_truncated": false
      },
      "response_received": true,
      "response_timing": {
        "content_chunks": 21,
        "controlled_output": null,
        "decode_tok_s": 156.6821051483209,
        "e2e_ms": 871.9521790008002,
        "effective_prefill_tok_s": 5442.377646323283,
        "generation_ms": 325.49983900025836,
        "mean_inter_token_latency_ms": 6.382349784318792,
        "mean_time_per_output_token_ms": 6.382349784318792,
        "output_token_source": "usage",
        "output_tokens": 52,
        "planned_context_tokens": 4096,
        "prompt_tokens": 2974,
        "reasoning_chunks": 0,
        "request_canary": {
          "capture_complete": true,
          "foreign_marker_count": 0,
          "marker": "ANVIL_REQ_29_00008",
          "marker_at_start": false,
          "observed_marker_count": 1,
          "passed": false
        },
        "request_index": 8,
        "time_to_first_output_ms": 546.4523400005419,
        "tpot_ms": 6.382349784318792,
        "ttft_ms": 546.4527100011765,
        "visible_content_truncated": false,
        "visible_generation_ms": 325.4994689996238
      }
    },
    {
      "elapsed_ms": 833.0627089999325,
      "error_message": "request canary validation failed",
      "error_message_truncated": false,
      "error_type": "ValueError",
      "failure_class": "request_canary_failed",
      "performance_eligible": false,
      "planned_context_tokens": 4096,
      "request_canary": {
        "capture_complete": true,
        "foreign_marker_count": 0,
        "marker": "ANVIL_REQ_29_00009",
        "marker_at_start": false,
        "observed_marker_count": 1,
        "passed": false
      },
      "request_index": 9,
      "response_metadata": {
        "content_chunks": 18,
        "finish_reasons": [
          "stop"
        ],
        "output_token_source": "usage",
        "reasoning_chunks": 0,
        "stream_terminal_observed": true,
        "usage": {
          "completion_tokens": 47,
          "prompt_tokens": 2972,
          "total_tokens": 3019
        },
        "visible_content_truncated": false
      },
      "response_received": true,
      "response_timing": {
        "content_chunks": 18,
        "controlled_output": null,
        "decode_tok_s": 158.70309979614578,
        "e2e_ms": 832.9450589990302,
        "effective_prefill_tok_s": 5472.332564175968,
        "generation_ms": 289.84941100134165,
        "mean_inter_token_latency_ms": 6.301074152203079,
        "mean_time_per_output_token_ms": 6.301074152203079,
        "output_token_source": "usage",
        "output_tokens": 47,
        "planned_context_tokens": 4096,
        "prompt_tokens": 2972,
        "reasoning_chunks": 0,
        "request_canary": {
          "capture_complete": true,
          "foreign_marker_count": 0,
          "marker": "ANVIL_REQ_29_00009",
          "marker_at_start": false,
          "observed_marker_count": 1,
          "passed": false
        },
        "request_index": 9,
        "time_to_first_output_ms": 543.0956479976885,
        "tpot_ms": 6.301074152203079,
        "ttft_ms": 543.0958079996344,
        "visible_content_truncated": false,
        "visible_generation_ms": 289.84925099939574
      }
    }
  ],
  "finished_at": "2026-09-08T05:24:55Z",
  "finished_at_unix_ns": 1788845095680588651,
  "gpu": "2x RTX PRO 6000 Blackwell Max-Q TP2 Linux",
  "manifest": null,
  "max_context_tokens": 393216,
  "max_tokens": 256,
  "measurement_protocol": "capacity-v3",
  "metric_population": {
    "aggregate_throughput_eligible": false,
    "attempted_requests": 10,
    "excluded_requests": 9,
    "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": 1,
      "e2e": 1,
      "effective_prefill_tok_s": 1,
      "tpot": 1,
      "ttft": 1
    },
    "performance_eligible_requests": 1,
    "successful_requests": 1
  },
  "metrics": {
    "completion_tokens_max": 36.0,
    "completion_tokens_mean": 36.0,
    "completion_tokens_mean_ci95_high": 36.0,
    "completion_tokens_mean_ci95_low": 36.0,
    "completion_tokens_min": 36.0,
    "completion_tokens_p25": 36.0,
    "completion_tokens_p50": 36.0,
    "completion_tokens_p75": 36.0,
    "completion_tokens_p90": 36.0,
    "completion_tokens_p95": 36.0,
    "completion_tokens_p99": 36.0,
    "completion_tokens_samples": 1,
    "completion_tokens_stddev": 0.0,
    "content_chunks": 15,
    "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": 156.90001645568154,
    "decode_tok_s_mean": 156.90001645568154,
    "decode_tok_s_mean_ci95_high": 156.90001645568154,
    "decode_tok_s_mean_ci95_low": 156.90001645568154,
    "decode_tok_s_min": 156.90001645568154,
    "decode_tok_s_p25": 156.90001645568154,
    "decode_tok_s_p50": 156.90001645568154,
    "decode_tok_s_p75": 156.90001645568154,
    "decode_tok_s_p90": 156.90001645568154,
    "decode_tok_s_p95": 156.90001645568154,
    "decode_tok_s_p99": 156.90001645568154,
    "decode_tok_s_samples": 1,
    "decode_tok_s_stddev": 0.0,
    "e2e_max_ms": 774.293022997881,
    "e2e_mean_ci95_high_ms": 774.293022997881,
    "e2e_mean_ci95_low_ms": 774.293022997881,
    "e2e_mean_ms": 774.293022997881,
    "e2e_min_ms": 774.293022997881,
    "e2e_p25_ms": 774.293022997881,
    "e2e_p50_ms": 774.293022997881,
    "e2e_p75_ms": 774.293022997881,
    "e2e_p90_ms": 774.293022997881,
    "e2e_p95_ms": 774.293022997881,
    "e2e_p99_ms": 774.293022997881,
    "e2e_samples": 1,
    "e2e_stddev_ms": 0.0,
    "effective_prefill_tok_s_max": 5402.551534759387,
    "effective_prefill_tok_s_mean": 5402.551534759387,
    "effective_prefill_tok_s_mean_ci95_high": 5402.551534759387,
    "effective_prefill_tok_s_mean_ci95_low": 5402.551534759387,
    "effective_prefill_tok_s_min": 5402.551534759387,
    "effective_prefill_tok_s_p25": 5402.551534759387,
    "effective_prefill_tok_s_p50": 5402.551534759387,
    "effective_prefill_tok_s_p75": 5402.551534759387,
    "effective_prefill_tok_s_p90": 5402.551534759387,
    "effective_prefill_tok_s_p95": 5402.551534759387,
    "effective_prefill_tok_s_p99": 5402.551534759387,
    "effective_prefill_tok_s_samples": 1,
    "effective_prefill_tok_s_stddev": 0.0,
    "generation_max_ms": 223.07199699935154,
    "generation_mean_ci95_high_ms": 223.07199699935154,
    "generation_mean_ci95_low_ms": 223.07199699935154,
    "generation_mean_ms": 223.07199699935154,
    "generation_min_ms": 223.07199699935154,
    "generation_p25_ms": 223.07199699935154,
    "generation_p50_ms": 223.07199699935154,
    "generation_p75_ms": 223.07199699935154,
    "generation_p90_ms": 223.07199699935154,
    "generation_p95_ms": 223.07199699935154,
    "generation_p99_ms": 223.07199699935154,
    "generation_samples": 1,
    "generation_stddev_ms": 0.0,
    "mean_inter_token_latency_ms_max": 6.373485628552901,
    "mean_inter_token_latency_ms_mean": 6.373485628552901,
    "mean_inter_token_latency_ms_mean_ci95_high": 6.373485628552901,
    "mean_inter_token_latency_ms_mean_ci95_low": 6.373485628552901,
    "mean_inter_token_latency_ms_min": 6.373485628552901,
    "mean_inter_token_latency_ms_p25": 6.373485628552901,
    "mean_inter_token_latency_ms_p50": 6.373485628552901,
    "mean_inter_token_latency_ms_p75": 6.373485628552901,
    "mean_inter_token_latency_ms_p90": 6.373485628552901,
    "mean_inter_token_latency_ms_p95": 6.373485628552901,
    "mean_inter_token_latency_ms_p99": 6.373485628552901,
    "mean_inter_token_latency_ms_samples": 1,
    "mean_inter_token_latency_ms_stddev": 0.0,
    "mean_time_per_output_token_max_ms": 6.373485628552901,
    "mean_time_per_output_token_mean_ci95_high_ms": 6.373485628552901,
    "mean_time_per_output_token_mean_ci95_low_ms": 6.373485628552901,
    "mean_time_per_output_token_mean_ms": 6.373485628552901,
    "mean_time_per_output_token_min_ms": 6.373485628552901,
    "mean_time_per_output_token_p25_ms": 6.373485628552901,
    "mean_time_per_output_token_p50_ms": 6.373485628552901,
    "mean_time_per_output_token_p75_ms": 6.373485628552901,
    "mean_time_per_output_token_p90_ms": 6.373485628552901,
    "mean_time_per_output_token_p95_ms": 6.373485628552901,
    "mean_time_per_output_token_p99_ms": 6.373485628552901,
    "mean_time_per_output_token_samples": 1,
    "mean_time_per_output_token_stddev_ms": 0.0,
    "output_token_sources": {
      "usage": 1
    },
    "output_tokens": 36,
    "prefix_cache_hit_avg": null,
    "prompt_token_samples": 1,
    "prompt_tokens": 2978,
    "prompt_tokens_max": 2978.0,
    "prompt_tokens_mean": 2978.0,
    "prompt_tokens_mean_ci95_high": 2978.0,
    "prompt_tokens_mean_ci95_low": 2978.0,
    "prompt_tokens_min": 2978.0,
    "prompt_tokens_p25": 2978.0,
    "prompt_tokens_p50": 2978.0,
    "prompt_tokens_p75": 2978.0,
    "prompt_tokens_p90": 2978.0,
    "prompt_tokens_p95": 2978.0,
    "prompt_tokens_p99": 2978.0,
    "prompt_tokens_samples": 1,
    "prompt_tokens_stddev": 0.0,
    "reasoning_chunks": 0,
    "request_canary_passed": 1,
    "request_canary_samples": 1,
    "throughput_performance_eligible": false,
    "throughput_tok_s": null,
    "time_to_first_output_max_ms": 551.2210259985295,
    "time_to_first_output_mean_ci95_high_ms": 551.2210259985295,
    "time_to_first_output_mean_ci95_low_ms": 551.2210259985295,
    "time_to_first_output_mean_ms": 551.2210259985295,
    "time_to_first_output_min_ms": 551.2210259985295,
    "time_to_first_output_p25_ms": 551.2210259985295,
    "time_to_first_output_p50_ms": 551.2210259985295,
    "time_to_first_output_p75_ms": 551.2210259985295,
    "time_to_first_output_p90_ms": 551.2210259985295,
    "time_to_first_output_p95_ms": 551.2210259985295,
    "time_to_first_output_p99_ms": 551.2210259985295,
    "time_to_first_output_samples": 1,
    "time_to_first_output_stddev_ms": 0.0,
    "tpot_max_ms": 6.373485628552901,
    "tpot_mean_ci95_high_ms": 6.373485628552901,
    "tpot_mean_ci95_low_ms": 6.373485628552901,
    "tpot_mean_ms": 6.373485628552901,
    "tpot_min_ms": 6.373485628552901,
    "tpot_p25_ms": 6.373485628552901,
    "tpot_p50_ms": 6.373485628552901,
    "tpot_p75_ms": 6.373485628552901,
    "tpot_p90_ms": 6.373485628552901,
    "tpot_p95_ms": 6.373485628552901,
    "tpot_p99_ms": 6.373485628552901,
    "tpot_samples": 1,
    "tpot_stddev_ms": 0.0,
    "ttft_max_ms": 551.2215460003063,
    "ttft_mean_ci95_high_ms": 551.2215460003063,
    "ttft_mean_ci95_low_ms": 551.2215460003063,
    "ttft_mean_ms": 551.2215460003063,
    "ttft_min_ms": 551.2215460003063,
    "ttft_p25_ms": 551.2215460003063,
    "ttft_p50_ms": 551.2215460003063,
    "ttft_p75_ms": 551.2215460003063,
    "ttft_p90_ms": 551.2215460003063,
    "ttft_p95_ms": 551.2215460003063,
    "ttft_p99_ms": 551.2215460003063,
    "ttft_samples": 1,
    "ttft_stddev_ms": 0.0
  },
  "model": "glm53-flash-ormandj-sglang-sm120-tp2-393k-c1-adaptive-mtp",
  "performance_eligible": false,
  "prompt_cache_mode": "unique",
  "request_canaries": true,
  "request_timings": [
    {
      "content_chunks": 15,
      "controlled_output": null,
      "decode_tok_s": 156.90001645568154,
      "e2e_ms": 774.293022997881,
      "effective_prefill_tok_s": 5402.551534759387,
      "generation_ms": 223.07199699935154,
      "mean_inter_token_latency_ms": 6.373485628552901,
      "mean_time_per_output_token_ms": 6.373485628552901,
      "output_token_source": "usage",
      "output_tokens": 36,
      "planned_context_tokens": 4096,
      "prompt_tokens": 2978,
      "reasoning_chunks": 0,
      "request_canary": {
        "capture_complete": true,
        "foreign_marker_count": 0,
        "marker": "ANVIL_REQ_29_00001",
        "marker_at_start": true,
        "observed_marker_count": 1,
        "passed": true
      },
      "request_index": 1,
      "time_to_first_output_ms": 551.2210259985295,
      "tpot_ms": 6.373485628552901,
      "ttft_ms": 551.2215460003063,
      "visible_content_truncated": false,
      "visible_generation_ms": 223.0714769975748
    }
  ],
  "requests": 10,
  "response_words": 0,
  "responses_received": 10,
  "run_id": "benchmark-20260908T052447Z",
  "schema": "anvil-serving.benchmark/v1",
  "serve_flags": {
    "no_thinking": false,
    "reasoning_effort": null,
    "shared_prefix_burst": false,
    "thinking_mode": "disabled"
  },
  "shared_prefix_tokens": 0,
  "source_recipe": null,
  "started_at": "2026-09-08T05:24:47Z",
  "started_at_unix_ns": 1788845087577106978,
  "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"
  },
  "wall_clock_ms": 8103.476033000334
}
