{
  "run_id": "support-extraction-v1-20260914114342",
  "started_at": "2026-09-14T11:43:42.637Z",
  "mode": "paired_run_verified",
  "prompt_profile": "long-cached",
  "prompt_version": "support-extraction-prompt-v1-long",
  "planned_cells": 180,
  "denominators": {
    "attempted": 180,
    "completed": 174,
    "completed_parsed": 174,
    "transport_failures_reported_separately": 6
  },
  "configurations": {
    "A": {
      "model_id": "anthropic/claude-sonnet-5",
      "attempted": 90,
      "completed": 87,
      "transport_failures": 3,
      "schema_pass": "84/87 (96.6%)",
      "task_acceptance": "81/90 (90.0%)",
      "schema_valid_task_fail": 3,
      "field_failures": {
        "urgency": 3
      },
      "failing_cases": {
        "case-20": [
          "urgency: expected \"normal\" got \"unknown\" (r1)",
          "urgency: expected \"normal\" got \"unknown\" (r2)",
          "urgency: expected \"normal\" got \"unknown\" (r3)"
        ]
      },
      "latency_ms": {
        "p50": 1563,
        "p95": 1828,
        "n": 87
      },
      "usage": {
        "prompt_tokens": 370206,
        "uncached_prompt_tokens": 3588,
        "cached_prompt_tokens": 362404,
        "cache_write_tokens": 4214,
        "completion_tokens": 8919,
        "reasoning_tokens_within_completion": 4980,
        "calls_with_cache_hit": "86/87",
        "finish_reasons": {
          "stop": 84,
          "length": 3
        }
      },
      "rates_usd_per_1m": {
        "input": 2,
        "output": 10,
        "cache_read": 0.2,
        "cache_write": 2.5
      },
      "calculated_list_cost_usd": 0.1794,
      "gateway_reported_charge_usd": 0.1794,
      "cost_per_acceptable_result_usd": 0.00221
    },
    "B": {
      "model_id": "openai/gpt-5.6-terra",
      "attempted": 90,
      "completed": 87,
      "transport_failures": 3,
      "schema_pass": "87/87 (100.0%)",
      "task_acceptance": "84/90 (93.3%)",
      "schema_valid_task_fail": 3,
      "field_failures": {
        "urgency": 3
      },
      "failing_cases": {
        "case-20": [
          "urgency: expected \"normal\" got \"unknown\" (r1)",
          "urgency: expected \"normal\" got \"unknown\" (r2)",
          "urgency: expected \"normal\" got \"unknown\" (r3)"
        ]
      },
      "latency_ms": {
        "p50": 1027,
        "p95": 1415,
        "n": 87
      },
      "usage": {
        "prompt_tokens": 234564,
        "uncached_prompt_tokens": 261,
        "cached_prompt_tokens": 230822,
        "cache_write_tokens": 3481,
        "completion_tokens": 3135,
        "reasoning_tokens_within_completion": 0,
        "calls_with_cache_hit": "86/87",
        "finish_reasons": {
          "stop": 87
        }
      },
      "rates_usd_per_1m": {
        "input": 2,
        "output": 12,
        "cache_read": 0.2,
        "cache_write": 2.5
      },
      "calculated_list_cost_usd": 0.093,
      "gateway_reported_charge_usd": 0.093,
      "cost_per_acceptable_result_usd": 0.00111
    }
  },
  "case_agreement": {
    "both_pass_all_reps": 27,
    "a_only_pass_all_reps": 0,
    "b_only_pass_all_reps": 1,
    "neither_pass_all_reps": 2,
    "unstable_across_reps": []
  },
  "label": "Paired-run provenance validated. Outputs are caller-supplied; their model origin is not independently verified. Results apply only to this synthetic fixture, not general quality or production safety."
}