{
  "base_url": "http://127.0.0.1:30114/v1",
  "clock_domain_id": null,
  "completed": 0,
  "concurrency": 1,
  "configuration_fingerprint": null,
  "context_distribution": {
    "4096": 10
  },
  "context_seed": 0,
  "context_tokens": 4096,
  "controlled_output_policy": "strict",
  "engine": "mlx-lm",
  "failed": 10,
  "failures": [
    {
      "controlled_output": {
        "capture_complete": true,
        "exact_adherence": false,
        "observed_code_words": 28,
        "observed_extra_words": 0,
        "passed": false,
        "policy": "strict",
        "requested_words": 32
      },
      "elapsed_ms": 4986.498749989551,
      "error_message": "controlled output validation failed",
      "error_message_truncated": false,
      "error_type": "ValueError",
      "failure_class": "controlled_output_failed",
      "performance_eligible": false,
      "planned_context_tokens": 4096,
      "request_canary": {
        "capture_complete": true,
        "foreign_marker_count": 0,
        "marker": "ANVIL_REQ_0_00000",
        "marker_at_start": true,
        "observed_marker_count": 1,
        "passed": true
      },
      "request_index": 0,
      "response_metadata": {
        "content_chunks": 40,
        "finish_reasons": [
          "stop"
        ],
        "output_token_source": "usage",
        "reasoning_chunks": 0,
        "stream_terminal_observed": true,
        "usage": {
          "completion_tokens": 41,
          "prompt_tokens": 3612,
          "total_tokens": 3653
        },
        "visible_content_truncated": false
      },
      "response_received": true,
      "response_timing": {
        "content_chunks": 40,
        "controlled_output": {
          "capture_complete": true,
          "exact_adherence": false,
          "observed_code_words": 28,
          "observed_extra_words": 0,
          "passed": false,
          "policy": "strict",
          "requested_words": 32
        },
        "decode_tok_s": 75.66859107078972,
        "e2e_ms": 4985.6685000122525,
        "effective_prefill_tok_s": 810.4019382174779,
        "generation_ms": 528.6209170008078,
        "mean_inter_token_latency_ms": 13.215522925020196,
        "mean_time_per_output_token_ms": 13.215522925020196,
        "output_token_source": "usage",
        "output_tokens": 41,
        "planned_context_tokens": 4096,
        "prompt_tokens": 3612,
        "reasoning_chunks": 0,
        "request_canary": {
          "capture_complete": true,
          "foreign_marker_count": 0,
          "marker": "ANVIL_REQ_0_00000",
          "marker_at_start": true,
          "observed_marker_count": 1,
          "passed": true
        },
        "request_index": 0,
        "time_to_first_output_ms": 4457.047583011445,
        "tpot_ms": 13.215522925020196,
        "ttft_ms": 4457.049707998522,
        "visible_content_truncated": false,
        "visible_generation_ms": 528.6187920137309
      }
    },
    {
      "controlled_output": {
        "capture_complete": true,
        "exact_adherence": false,
        "observed_code_words": 28,
        "observed_extra_words": 0,
        "passed": false,
        "policy": "strict",
        "requested_words": 32
      },
      "elapsed_ms": 4800.355292012682,
      "error_message": "controlled output validation failed",
      "error_message_truncated": false,
      "error_type": "ValueError",
      "failure_class": "controlled_output_failed",
      "performance_eligible": false,
      "planned_context_tokens": 4096,
      "request_canary": {
        "capture_complete": true,
        "foreign_marker_count": 0,
        "marker": "ANVIL_REQ_0_00001",
        "marker_at_start": true,
        "observed_marker_count": 1,
        "passed": true
      },
      "request_index": 1,
      "response_metadata": {
        "content_chunks": 40,
        "finish_reasons": [
          "stop"
        ],
        "output_token_source": "usage",
        "reasoning_chunks": 0,
        "stream_terminal_observed": true,
        "usage": {
          "completion_tokens": 41,
          "prompt_tokens": 3615,
          "total_tokens": 3656
        },
        "visible_content_truncated": false
      },
      "response_received": true,
      "response_timing": {
        "content_chunks": 40,
        "controlled_output": {
          "capture_complete": true,
          "exact_adherence": false,
          "observed_code_words": 28,
          "observed_extra_words": 0,
          "passed": false,
          "policy": "strict",
          "requested_words": 32
        },
        "decode_tok_s": 75.9466152261383,
        "e2e_ms": 4799.854458979098,
        "effective_prefill_tok_s": 845.9764278438403,
        "generation_ms": 526.6857499955222,
        "mean_inter_token_latency_ms": 13.167143749888055,
        "mean_time_per_output_token_ms": 13.167143749888055,
        "output_token_source": "usage",
        "output_tokens": 41,
        "planned_context_tokens": 4096,
        "prompt_tokens": 3615,
        "reasoning_chunks": 0,
        "request_canary": {
          "capture_complete": true,
          "foreign_marker_count": 0,
          "marker": "ANVIL_REQ_0_00001",
          "marker_at_start": true,
          "observed_marker_count": 1,
          "passed": true
        },
        "request_index": 1,
        "time_to_first_output_ms": 4273.168708983576,
        "tpot_ms": 13.167143749888055,
        "ttft_ms": 4273.169583990239,
        "visible_content_truncated": false,
        "visible_generation_ms": 526.6848749888595
      }
    },
    {
      "controlled_output": {
        "capture_complete": true,
        "exact_adherence": false,
        "observed_code_words": 28,
        "observed_extra_words": 0,
        "passed": false,
        "policy": "strict",
        "requested_words": 32
      },
      "elapsed_ms": 4842.328458005795,
      "error_message": "controlled output validation failed",
      "error_message_truncated": false,
      "error_type": "ValueError",
      "failure_class": "controlled_output_failed",
      "performance_eligible": false,
      "planned_context_tokens": 4096,
      "request_canary": {
        "capture_complete": true,
        "foreign_marker_count": 0,
        "marker": "ANVIL_REQ_0_00002",
        "marker_at_start": true,
        "observed_marker_count": 1,
        "passed": true
      },
      "request_index": 2,
      "response_metadata": {
        "content_chunks": 40,
        "finish_reasons": [
          "stop"
        ],
        "output_token_source": "usage",
        "reasoning_chunks": 0,
        "stream_terminal_observed": true,
        "usage": {
          "completion_tokens": 41,
          "prompt_tokens": 3616,
          "total_tokens": 3657
        },
        "visible_content_truncated": false
      },
      "response_received": true,
      "response_timing": {
        "content_chunks": 40,
        "controlled_output": {
          "capture_complete": true,
          "exact_adherence": false,
          "observed_code_words": 28,
          "observed_extra_words": 0,
          "passed": false,
          "policy": "strict",
          "requested_words": 32
        },
        "decode_tok_s": 75.80352331633826,
        "e2e_ms": 4841.999167023459,
        "effective_prefill_tok_s": 838.1391883194698,
        "generation_ms": 527.679958002409,
        "mean_inter_token_latency_ms": 13.191998950060224,
        "mean_time_per_output_token_ms": 13.191998950060224,
        "output_token_source": "usage",
        "output_tokens": 41,
        "planned_context_tokens": 4096,
        "prompt_tokens": 3616,
        "reasoning_chunks": 0,
        "request_canary": {
          "capture_complete": true,
          "foreign_marker_count": 0,
          "marker": "ANVIL_REQ_0_00002",
          "marker_at_start": true,
          "observed_marker_count": 1,
          "passed": true
        },
        "request_index": 2,
        "time_to_first_output_ms": 4314.3192090210505,
        "tpot_ms": 13.191998950060224,
        "ttft_ms": 4314.319459022954,
        "visible_content_truncated": false,
        "visible_generation_ms": 527.6797080005053
      }
    },
    {
      "controlled_output": {
        "capture_complete": true,
        "exact_adherence": false,
        "observed_code_words": 28,
        "observed_extra_words": 0,
        "passed": false,
        "policy": "strict",
        "requested_words": 32
      },
      "elapsed_ms": 4924.006042012479,
      "error_message": "controlled output validation failed",
      "error_message_truncated": false,
      "error_type": "ValueError",
      "failure_class": "controlled_output_failed",
      "performance_eligible": false,
      "planned_context_tokens": 4096,
      "request_canary": {
        "capture_complete": true,
        "foreign_marker_count": 0,
        "marker": "ANVIL_REQ_0_00003",
        "marker_at_start": true,
        "observed_marker_count": 1,
        "passed": true
      },
      "request_index": 3,
      "response_metadata": {
        "content_chunks": 40,
        "finish_reasons": [
          "stop"
        ],
        "output_token_source": "usage",
        "reasoning_chunks": 0,
        "stream_terminal_observed": true,
        "usage": {
          "completion_tokens": 41,
          "prompt_tokens": 3616,
          "total_tokens": 3657
        },
        "visible_content_truncated": false
      },
      "response_received": true,
      "response_timing": {
        "content_chunks": 40,
        "controlled_output": {
          "capture_complete": true,
          "exact_adherence": false,
          "observed_code_words": 28,
          "observed_extra_words": 0,
          "passed": false,
          "policy": "strict",
          "requested_words": 32
        },
        "decode_tok_s": 76.32045108266675,
        "e2e_ms": 4923.536999995122,
        "effective_prefill_tok_s": 821.9244560895537,
        "generation_ms": 524.105916992994,
        "mean_inter_token_latency_ms": 13.10264792482485,
        "mean_time_per_output_token_ms": 13.10264792482485,
        "output_token_source": "usage",
        "output_tokens": 41,
        "planned_context_tokens": 4096,
        "prompt_tokens": 3616,
        "reasoning_chunks": 0,
        "request_canary": {
          "capture_complete": true,
          "foreign_marker_count": 0,
          "marker": "ANVIL_REQ_0_00003",
          "marker_at_start": true,
          "observed_marker_count": 1,
          "passed": true
        },
        "request_index": 3,
        "time_to_first_output_ms": 4399.431083002128,
        "tpot_ms": 13.10264792482485,
        "ttft_ms": 4399.431707977783,
        "visible_content_truncated": false,
        "visible_generation_ms": 524.1052920173388
      }
    },
    {
      "controlled_output": {
        "capture_complete": true,
        "exact_adherence": false,
        "observed_code_words": 28,
        "observed_extra_words": 0,
        "passed": false,
        "policy": "strict",
        "requested_words": 32
      },
      "elapsed_ms": 5199.941625003703,
      "error_message": "controlled output validation failed",
      "error_message_truncated": false,
      "error_type": "ValueError",
      "failure_class": "controlled_output_failed",
      "performance_eligible": false,
      "planned_context_tokens": 4096,
      "request_canary": {
        "capture_complete": true,
        "foreign_marker_count": 0,
        "marker": "ANVIL_REQ_0_00004",
        "marker_at_start": true,
        "observed_marker_count": 1,
        "passed": true
      },
      "request_index": 4,
      "response_metadata": {
        "content_chunks": 40,
        "finish_reasons": [
          "stop"
        ],
        "output_token_source": "usage",
        "reasoning_chunks": 0,
        "stream_terminal_observed": true,
        "usage": {
          "completion_tokens": 41,
          "prompt_tokens": 3618,
          "total_tokens": 3659
        },
        "visible_content_truncated": false
      },
      "response_received": true,
      "response_timing": {
        "content_chunks": 40,
        "controlled_output": {
          "capture_complete": true,
          "exact_adherence": false,
          "observed_code_words": 28,
          "observed_extra_words": 0,
          "passed": false,
          "policy": "strict",
          "requested_words": 32
        },
        "decode_tok_s": 71.36820590571867,
        "e2e_ms": 5199.6637909905985,
        "effective_prefill_tok_s": 779.8775007112449,
        "generation_ms": 560.4736660025083,
        "mean_inter_token_latency_ms": 14.011841650062706,
        "mean_time_per_output_token_ms": 14.011841650062706,
        "output_token_source": "usage",
        "output_tokens": 41,
        "planned_context_tokens": 4096,
        "prompt_tokens": 3618,
        "reasoning_chunks": 0,
        "request_canary": {
          "capture_complete": true,
          "foreign_marker_count": 0,
          "marker": "ANVIL_REQ_0_00004",
          "marker_at_start": true,
          "observed_marker_count": 1,
          "passed": true
        },
        "request_index": 4,
        "time_to_first_output_ms": 4639.19012498809,
        "tpot_ms": 14.011841650062706,
        "ttft_ms": 4639.190457994118,
        "visible_content_truncated": false,
        "visible_generation_ms": 560.4733329964802
      }
    },
    {
      "controlled_output": {
        "capture_complete": true,
        "exact_adherence": false,
        "observed_code_words": 28,
        "observed_extra_words": 0,
        "passed": false,
        "policy": "strict",
        "requested_words": 32
      },
      "elapsed_ms": 5430.405208986485,
      "error_message": "controlled output validation failed",
      "error_message_truncated": false,
      "error_type": "ValueError",
      "failure_class": "controlled_output_failed",
      "performance_eligible": false,
      "planned_context_tokens": 4096,
      "request_canary": {
        "capture_complete": true,
        "foreign_marker_count": 0,
        "marker": "ANVIL_REQ_0_00005",
        "marker_at_start": true,
        "observed_marker_count": 1,
        "passed": true
      },
      "request_index": 5,
      "response_metadata": {
        "content_chunks": 40,
        "finish_reasons": [
          "stop"
        ],
        "output_token_source": "usage",
        "reasoning_chunks": 0,
        "stream_terminal_observed": true,
        "usage": {
          "completion_tokens": 41,
          "prompt_tokens": 3619,
          "total_tokens": 3660
        },
        "visible_content_truncated": false
      },
      "response_received": true,
      "response_timing": {
        "content_chunks": 40,
        "controlled_output": {
          "capture_complete": true,
          "exact_adherence": false,
          "observed_code_words": 28,
          "observed_extra_words": 0,
          "passed": false,
          "policy": "strict",
          "requested_words": 32
        },
        "decode_tok_s": 74.37623330897956,
        "e2e_ms": 5430.252124991966,
        "effective_prefill_tok_s": 739.7118049723662,
        "generation_ms": 537.8062079835217,
        "mean_inter_token_latency_ms": 13.445155199588044,
        "mean_time_per_output_token_ms": 13.445155199588044,
        "output_token_source": "usage",
        "output_tokens": 41,
        "planned_context_tokens": 4096,
        "prompt_tokens": 3619,
        "reasoning_chunks": 0,
        "request_canary": {
          "capture_complete": true,
          "foreign_marker_count": 0,
          "marker": "ANVIL_REQ_0_00005",
          "marker_at_start": true,
          "observed_marker_count": 1,
          "passed": true
        },
        "request_index": 5,
        "time_to_first_output_ms": 4892.445917008445,
        "tpot_ms": 13.445155199588044,
        "ttft_ms": 4892.446375015425,
        "visible_content_truncated": false,
        "visible_generation_ms": 537.8057499765418
      }
    },
    {
      "controlled_output": {
        "capture_complete": true,
        "exact_adherence": false,
        "observed_code_words": 28,
        "observed_extra_words": 0,
        "passed": false,
        "policy": "strict",
        "requested_words": 32
      },
      "elapsed_ms": 5444.653708022088,
      "error_message": "controlled output validation failed",
      "error_message_truncated": false,
      "error_type": "ValueError",
      "failure_class": "controlled_output_failed",
      "performance_eligible": false,
      "planned_context_tokens": 4096,
      "request_canary": {
        "capture_complete": true,
        "foreign_marker_count": 0,
        "marker": "ANVIL_REQ_0_00006",
        "marker_at_start": true,
        "observed_marker_count": 1,
        "passed": true
      },
      "request_index": 6,
      "response_metadata": {
        "content_chunks": 40,
        "finish_reasons": [
          "stop"
        ],
        "output_token_source": "usage",
        "reasoning_chunks": 0,
        "stream_terminal_observed": true,
        "usage": {
          "completion_tokens": 41,
          "prompt_tokens": 3621,
          "total_tokens": 3662
        },
        "visible_content_truncated": false
      },
      "response_received": true,
      "response_timing": {
        "content_chunks": 40,
        "controlled_output": {
          "capture_complete": true,
          "exact_adherence": false,
          "observed_code_words": 28,
          "observed_extra_words": 0,
          "passed": false,
          "policy": "strict",
          "requested_words": 32
        },
        "decode_tok_s": 74.11893442183658,
        "e2e_ms": 5444.417708000401,
        "effective_prefill_tok_s": 738.26474936519,
        "generation_ms": 539.6731660002843,
        "mean_inter_token_latency_ms": 13.491829150007106,
        "mean_time_per_output_token_ms": 13.491829150007106,
        "output_token_source": "usage",
        "output_tokens": 41,
        "planned_context_tokens": 4096,
        "prompt_tokens": 3621,
        "reasoning_chunks": 0,
        "request_canary": {
          "capture_complete": true,
          "foreign_marker_count": 0,
          "marker": "ANVIL_REQ_0_00006",
          "marker_at_start": true,
          "observed_marker_count": 1,
          "passed": true
        },
        "request_index": 6,
        "time_to_first_output_ms": 4904.744542000117,
        "tpot_ms": 13.491829150007106,
        "ttft_ms": 4904.744917002972,
        "visible_content_truncated": false,
        "visible_generation_ms": 539.6727909974288
      }
    },
    {
      "controlled_output": {
        "capture_complete": true,
        "exact_adherence": false,
        "observed_code_words": 28,
        "observed_extra_words": 0,
        "passed": false,
        "policy": "strict",
        "requested_words": 32
      },
      "elapsed_ms": 5412.152250006329,
      "error_message": "controlled output validation failed",
      "error_message_truncated": false,
      "error_type": "ValueError",
      "failure_class": "controlled_output_failed",
      "performance_eligible": false,
      "planned_context_tokens": 4096,
      "request_canary": {
        "capture_complete": true,
        "foreign_marker_count": 0,
        "marker": "ANVIL_REQ_0_00007",
        "marker_at_start": true,
        "observed_marker_count": 1,
        "passed": true
      },
      "request_index": 7,
      "response_metadata": {
        "content_chunks": 40,
        "finish_reasons": [
          "stop"
        ],
        "output_token_source": "usage",
        "reasoning_chunks": 0,
        "stream_terminal_observed": true,
        "usage": {
          "completion_tokens": 41,
          "prompt_tokens": 3622,
          "total_tokens": 3663
        },
        "visible_content_truncated": false
      },
      "response_received": true,
      "response_timing": {
        "content_chunks": 40,
        "controlled_output": {
          "capture_complete": true,
          "exact_adherence": false,
          "observed_code_words": 28,
          "observed_extra_words": 0,
          "passed": false,
          "policy": "strict",
          "requested_words": 32
        },
        "decode_tok_s": 75.73932229405881,
        "e2e_ms": 5411.919041973306,
        "effective_prefill_tok_s": 741.6368580562029,
        "generation_ms": 528.1272499996703,
        "mean_inter_token_latency_ms": 13.203181249991758,
        "mean_time_per_output_token_ms": 13.203181249991758,
        "output_token_source": "usage",
        "output_tokens": 41,
        "planned_context_tokens": 4096,
        "prompt_tokens": 3622,
        "reasoning_chunks": 0,
        "request_canary": {
          "capture_complete": true,
          "foreign_marker_count": 0,
          "marker": "ANVIL_REQ_0_00007",
          "marker_at_start": true,
          "observed_marker_count": 1,
          "passed": true
        },
        "request_index": 7,
        "time_to_first_output_ms": 4883.791791973636,
        "tpot_ms": 13.203181249991758,
        "ttft_ms": 4883.792207983788,
        "visible_content_truncated": false,
        "visible_generation_ms": 528.1268339895178
      }
    },
    {
      "controlled_output": {
        "capture_complete": true,
        "exact_adherence": false,
        "observed_code_words": 28,
        "observed_extra_words": 0,
        "passed": false,
        "policy": "strict",
        "requested_words": 32
      },
      "elapsed_ms": 5387.237915972946,
      "error_message": "controlled output validation failed",
      "error_message_truncated": false,
      "error_type": "ValueError",
      "failure_class": "controlled_output_failed",
      "performance_eligible": false,
      "planned_context_tokens": 4096,
      "request_canary": {
        "capture_complete": true,
        "foreign_marker_count": 0,
        "marker": "ANVIL_REQ_0_00008",
        "marker_at_start": true,
        "observed_marker_count": 1,
        "passed": true
      },
      "request_index": 8,
      "response_metadata": {
        "content_chunks": 40,
        "finish_reasons": [
          "stop"
        ],
        "output_token_source": "usage",
        "reasoning_chunks": 0,
        "stream_terminal_observed": true,
        "usage": {
          "completion_tokens": 41,
          "prompt_tokens": 3622,
          "total_tokens": 3663
        },
        "visible_content_truncated": false
      },
      "response_received": true,
      "response_timing": {
        "content_chunks": 40,
        "controlled_output": {
          "capture_complete": true,
          "exact_adherence": false,
          "observed_code_words": 28,
          "observed_extra_words": 0,
          "passed": false,
          "policy": "strict",
          "requested_words": 32
        },
        "decode_tok_s": 74.95750841537705,
        "e2e_ms": 5386.974208988249,
        "effective_prefill_tok_s": 746.2904078636982,
        "generation_ms": 533.6356670013629,
        "mean_inter_token_latency_ms": 13.340891675034072,
        "mean_time_per_output_token_ms": 13.340891675034072,
        "output_token_source": "usage",
        "output_tokens": 41,
        "planned_context_tokens": 4096,
        "prompt_tokens": 3622,
        "reasoning_chunks": 0,
        "request_canary": {
          "capture_complete": true,
          "foreign_marker_count": 0,
          "marker": "ANVIL_REQ_0_00008",
          "marker_at_start": true,
          "observed_marker_count": 1,
          "passed": true
        },
        "request_index": 8,
        "time_to_first_output_ms": 4853.338541986886,
        "tpot_ms": 13.340891675034072,
        "ttft_ms": 4853.338708984666,
        "visible_content_truncated": false,
        "visible_generation_ms": 533.6355000035837
      }
    },
    {
      "controlled_output": {
        "capture_complete": true,
        "exact_adherence": false,
        "observed_code_words": 28,
        "observed_extra_words": 0,
        "passed": false,
        "policy": "strict",
        "requested_words": 32
      },
      "elapsed_ms": 5338.004165998427,
      "error_message": "controlled output validation failed",
      "error_message_truncated": false,
      "error_type": "ValueError",
      "failure_class": "controlled_output_failed",
      "performance_eligible": false,
      "planned_context_tokens": 4096,
      "request_canary": {
        "capture_complete": true,
        "foreign_marker_count": 0,
        "marker": "ANVIL_REQ_0_00009",
        "marker_at_start": true,
        "observed_marker_count": 1,
        "passed": true
      },
      "request_index": 9,
      "response_metadata": {
        "content_chunks": 40,
        "finish_reasons": [
          "stop"
        ],
        "output_token_source": "usage",
        "reasoning_chunks": 0,
        "stream_terminal_observed": true,
        "usage": {
          "completion_tokens": 41,
          "prompt_tokens": 3622,
          "total_tokens": 3663
        },
        "visible_content_truncated": false
      },
      "response_received": true,
      "response_timing": {
        "content_chunks": 40,
        "controlled_output": {
          "capture_complete": true,
          "exact_adherence": false,
          "observed_code_words": 28,
          "observed_extra_words": 0,
          "passed": false,
          "policy": "strict",
          "requested_words": 32
        },
        "decode_tok_s": 75.38795347622641,
        "e2e_ms": 5337.515291990712,
        "effective_prefill_tok_s": 753.4960162923386,
        "generation_ms": 530.5887499998789,
        "mean_inter_token_latency_ms": 13.264718749996973,
        "mean_time_per_output_token_ms": 13.264718749996973,
        "output_token_source": "usage",
        "output_tokens": 41,
        "planned_context_tokens": 4096,
        "prompt_tokens": 3622,
        "reasoning_chunks": 0,
        "request_canary": {
          "capture_complete": true,
          "foreign_marker_count": 0,
          "marker": "ANVIL_REQ_0_00009",
          "marker_at_start": true,
          "observed_marker_count": 1,
          "passed": true
        },
        "request_index": 9,
        "time_to_first_output_ms": 4806.926541990833,
        "tpot_ms": 13.264718749996973,
        "ttft_ms": 4806.927333993372,
        "visible_content_truncated": false,
        "visible_generation_ms": 530.5879579973407
      }
    }
  ],
  "finished_at": "2026-09-08T09:31:55Z",
  "finished_at_unix_ns": 1788859915283263000,
  "gpu": null,
  "manifest": null,
  "max_context_tokens": 8192,
  "max_tokens": 256,
  "measurement_protocol": "capacity-v3",
  "metric_population": {
    "aggregate_throughput_eligible": false,
    "attempted_requests": 10,
    "excluded_requests": 10,
    "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": 0,
      "e2e": 0,
      "effective_prefill_tok_s": 0,
      "tpot": 0,
      "ttft": 0
    },
    "performance_eligible_requests": 0,
    "successful_requests": 0
  },
  "metrics": {
    "completion_tokens_max": null,
    "completion_tokens_mean": null,
    "completion_tokens_mean_ci95_high": null,
    "completion_tokens_mean_ci95_low": null,
    "completion_tokens_min": null,
    "completion_tokens_p25": null,
    "completion_tokens_p50": null,
    "completion_tokens_p75": null,
    "completion_tokens_p90": null,
    "completion_tokens_p95": null,
    "completion_tokens_p99": null,
    "completion_tokens_samples": 0,
    "completion_tokens_stddev": null,
    "content_chunks": 0,
    "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": null,
    "decode_tok_s_mean": null,
    "decode_tok_s_mean_ci95_high": null,
    "decode_tok_s_mean_ci95_low": null,
    "decode_tok_s_min": null,
    "decode_tok_s_p25": null,
    "decode_tok_s_p50": null,
    "decode_tok_s_p75": null,
    "decode_tok_s_p90": null,
    "decode_tok_s_p95": null,
    "decode_tok_s_p99": null,
    "decode_tok_s_samples": 0,
    "decode_tok_s_stddev": null,
    "e2e_max_ms": null,
    "e2e_mean_ci95_high_ms": null,
    "e2e_mean_ci95_low_ms": null,
    "e2e_mean_ms": null,
    "e2e_min_ms": null,
    "e2e_p25_ms": null,
    "e2e_p50_ms": null,
    "e2e_p75_ms": null,
    "e2e_p90_ms": null,
    "e2e_p95_ms": null,
    "e2e_p99_ms": null,
    "e2e_samples": 0,
    "e2e_stddev_ms": null,
    "effective_prefill_tok_s_max": null,
    "effective_prefill_tok_s_mean": null,
    "effective_prefill_tok_s_mean_ci95_high": null,
    "effective_prefill_tok_s_mean_ci95_low": null,
    "effective_prefill_tok_s_min": null,
    "effective_prefill_tok_s_p25": null,
    "effective_prefill_tok_s_p50": null,
    "effective_prefill_tok_s_p75": null,
    "effective_prefill_tok_s_p90": null,
    "effective_prefill_tok_s_p95": null,
    "effective_prefill_tok_s_p99": null,
    "effective_prefill_tok_s_samples": 0,
    "effective_prefill_tok_s_stddev": null,
    "generation_max_ms": null,
    "generation_mean_ci95_high_ms": null,
    "generation_mean_ci95_low_ms": null,
    "generation_mean_ms": null,
    "generation_min_ms": null,
    "generation_p25_ms": null,
    "generation_p50_ms": null,
    "generation_p75_ms": null,
    "generation_p90_ms": null,
    "generation_p95_ms": null,
    "generation_p99_ms": null,
    "generation_samples": 0,
    "generation_stddev_ms": null,
    "mean_inter_token_latency_ms_max": null,
    "mean_inter_token_latency_ms_mean": null,
    "mean_inter_token_latency_ms_mean_ci95_high": null,
    "mean_inter_token_latency_ms_mean_ci95_low": null,
    "mean_inter_token_latency_ms_min": null,
    "mean_inter_token_latency_ms_p25": null,
    "mean_inter_token_latency_ms_p50": null,
    "mean_inter_token_latency_ms_p75": null,
    "mean_inter_token_latency_ms_p90": null,
    "mean_inter_token_latency_ms_p95": null,
    "mean_inter_token_latency_ms_p99": null,
    "mean_inter_token_latency_ms_samples": 0,
    "mean_inter_token_latency_ms_stddev": null,
    "mean_time_per_output_token_max_ms": null,
    "mean_time_per_output_token_mean_ci95_high_ms": null,
    "mean_time_per_output_token_mean_ci95_low_ms": null,
    "mean_time_per_output_token_mean_ms": null,
    "mean_time_per_output_token_min_ms": null,
    "mean_time_per_output_token_p25_ms": null,
    "mean_time_per_output_token_p50_ms": null,
    "mean_time_per_output_token_p75_ms": null,
    "mean_time_per_output_token_p90_ms": null,
    "mean_time_per_output_token_p95_ms": null,
    "mean_time_per_output_token_p99_ms": null,
    "mean_time_per_output_token_samples": 0,
    "mean_time_per_output_token_stddev_ms": null,
    "output_token_sources": {},
    "output_tokens": 0,
    "prefix_cache_hit_avg": null,
    "prompt_token_samples": 0,
    "prompt_tokens": null,
    "prompt_tokens_max": null,
    "prompt_tokens_mean": null,
    "prompt_tokens_mean_ci95_high": null,
    "prompt_tokens_mean_ci95_low": null,
    "prompt_tokens_min": null,
    "prompt_tokens_p25": null,
    "prompt_tokens_p50": null,
    "prompt_tokens_p75": null,
    "prompt_tokens_p90": null,
    "prompt_tokens_p95": null,
    "prompt_tokens_p99": null,
    "prompt_tokens_samples": 0,
    "prompt_tokens_stddev": null,
    "reasoning_chunks": 0,
    "request_canary_passed": 0,
    "request_canary_samples": 0,
    "throughput_performance_eligible": false,
    "throughput_tok_s": null,
    "time_to_first_output_max_ms": null,
    "time_to_first_output_mean_ci95_high_ms": null,
    "time_to_first_output_mean_ci95_low_ms": null,
    "time_to_first_output_mean_ms": null,
    "time_to_first_output_min_ms": null,
    "time_to_first_output_p25_ms": null,
    "time_to_first_output_p50_ms": null,
    "time_to_first_output_p75_ms": null,
    "time_to_first_output_p90_ms": null,
    "time_to_first_output_p95_ms": null,
    "time_to_first_output_p99_ms": null,
    "time_to_first_output_samples": 0,
    "time_to_first_output_stddev_ms": null,
    "tpot_max_ms": null,
    "tpot_mean_ci95_high_ms": null,
    "tpot_mean_ci95_low_ms": null,
    "tpot_mean_ms": null,
    "tpot_min_ms": null,
    "tpot_p25_ms": null,
    "tpot_p50_ms": null,
    "tpot_p75_ms": null,
    "tpot_p90_ms": null,
    "tpot_p95_ms": null,
    "tpot_p99_ms": null,
    "tpot_samples": 0,
    "tpot_stddev_ms": null,
    "ttft_max_ms": null,
    "ttft_mean_ci95_high_ms": null,
    "ttft_mean_ci95_low_ms": null,
    "ttft_mean_ms": null,
    "ttft_min_ms": null,
    "ttft_p25_ms": null,
    "ttft_p50_ms": null,
    "ttft_p75_ms": null,
    "ttft_p90_ms": null,
    "ttft_p95_ms": null,
    "ttft_p99_ms": null,
    "ttft_samples": 0,
    "ttft_stddev_ms": null
  },
  "model": "<model-cache>/models--mlx-community--Qwen3.5-9B-4bit/snapshots/8b2b98c00a6b4d291155e4890773ca8f769aee53",
  "performance_eligible": false,
  "prompt_cache_mode": "unique",
  "request_canaries": true,
  "request_timings": [],
  "requests": 10,
  "response_words": 32,
  "responses_received": 10,
  "run_id": "benchmark-20260908T093103Z",
  "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": "2026-09-08T09:31:03Z",
  "started_at_unix_ns": 1788859863516055000,
  "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": 51766.96416598861
}
