{
  "generatedAt": "2026-08-19T16:21:48.868Z",
  "cliVersion": "postil 0.8.26",
  "fixtureCorpusSha256": "8e4c2cb9ad5a7efdfe6a875566d20133e905155b6f693a873595adf6c069e065",
  "evaluatorSha256": "ebc72a8d7d60c08bf79759a0dfc5dbf821a9e6f79f4503c08ea9660dddece4d7",
  "benchmarkedBinarySha256": "37d0c7ef4b49ab27d55697e074f96c600025f80869bff9cb9344028b815f6d0b",
  "defectCases": 57,
  "cleanCases": 13,
  "models": [
    {
      "id": "openai/gpt-5.6-luna",
      "detectionRate": 0.9649122807017544,
      "gateVerdictCorrectness": 0.9,
      "falsePositives": 0,
      "cleanCasesSilent": 13,
      "cleanCases": 13,
      "unscoredCases": 0,
      "casesRun": 70,
      "totalCostUsd": 0.06282649,
      "meanCostUsdPerReview": 0.0008975212857142856,
      "latencyMsP50": 9318,
      "latencyMsP95": 21432,
      "pinnedProviderContract": {
        "provider": "Azure",
        "runs": 3,
        "detectionRate": 0.8947368421052632,
        "gateVerdictCorrectness": 0.8552631578947368,
        "falsePositives": 0,
        "totalCostUsd": 0.0555,
        "latencyMsP95": 17559
      }
    },
    {
      "id": "openai/gpt-5.6-terra",
      "detectionRate": 0.9473684210526315,
      "gateVerdictCorrectness": 0.8571428571428571,
      "falsePositives": 0,
      "cleanCasesSilent": 13,
      "cleanCases": 13,
      "unscoredCases": 0,
      "casesRun": 70,
      "totalCostUsd": 0.64024169,
      "meanCostUsdPerReview": 0.009146309857142857,
      "latencyMsP50": 8230,
      "latencyMsP95": 17603
    },
    {
      "id": "moonshotai/kimi-k2.7-code",
      "detectionRate": 0.9473684210526315,
      "gateVerdictCorrectness": 0.8260869565217391,
      "falsePositives": 0,
      "cleanCasesSilent": 13,
      "cleanCases": 13,
      "unscoredCases": 1,
      "casesRun": 70,
      "totalCostUsd": 0.51079418,
      "meanCostUsdPerReview": 0.007297059714285714,
      "latencyMsP50": 5225,
      "latencyMsP95": 18147,
      "pinnedProviderContract": {
        "provider": "CoreWeave",
        "runs": 1,
        "detectionRate": 0.9473684210526315,
        "gateVerdictCorrectness": 0.8550724637681159,
        "falsePositives": 0,
        "totalCostUsd": 0.5046,
        "latencyMsP95": 58200
      }
    },
    {
      "id": "z-ai/glm-5.2",
      "detectionRate": 0.9473684210526315,
      "gateVerdictCorrectness": 0.7681159420289855,
      "falsePositives": 3,
      "cleanCasesSilent": 13,
      "cleanCases": 13,
      "unscoredCases": 1,
      "casesRun": 70,
      "totalCostUsd": 0.63287166,
      "meanCostUsdPerReview": 0.009041023714285714,
      "latencyMsP50": 10272,
      "latencyMsP95": 62096
    },
    {
      "id": "openai/gpt-5.6-luna-pro",
      "detectionRate": 0.9298245614035088,
      "gateVerdictCorrectness": 0.8571428571428571,
      "falsePositives": 0,
      "cleanCasesSilent": 13,
      "cleanCases": 13,
      "unscoredCases": 0,
      "casesRun": 70,
      "totalCostUsd": 0.30326815,
      "meanCostUsdPerReview": 0.004332402142857143,
      "latencyMsP50": 27396,
      "latencyMsP95": 38836
    },
    {
      "id": "moonshotai/kimi-k2.6",
      "detectionRate": 0.9122807017543859,
      "gateVerdictCorrectness": 0.8428571428571429,
      "falsePositives": 1,
      "cleanCasesSilent": 13,
      "cleanCases": 13,
      "unscoredCases": 0,
      "casesRun": 70,
      "totalCostUsd": 1.60869745,
      "meanCostUsdPerReview": 0.022981392142857143,
      "latencyMsP50": 22231,
      "latencyMsP95": 57789
    },
    {
      "id": "google/gemma-4-31b-it",
      "detectionRate": 0.9122807017543859,
      "gateVerdictCorrectness": 0.7794117647058824,
      "falsePositives": 3,
      "cleanCasesSilent": 10,
      "cleanCases": 13,
      "unscoredCases": 2,
      "casesRun": 70,
      "totalCostUsd": 0.2846151,
      "meanCostUsdPerReview": 0.00406593,
      "latencyMsP50": 1636,
      "latencyMsP95": 11732
    },
    {
      "id": "qwen/qwen3.8-27b",
      "detectionRate": 0.9122807017543859,
      "gateVerdictCorrectness": 0.7391304347826086,
      "falsePositives": 1,
      "cleanCasesSilent": 13,
      "cleanCases": 13,
      "unscoredCases": 1,
      "casesRun": 70,
      "totalCostUsd": 0.98161095,
      "meanCostUsdPerReview": 0.014023013571428572,
      "latencyMsP50": 22766,
      "latencyMsP95": 67506
    },
    {
      "id": "mistralai/mistral-small-3.2-24b-instruct",
      "detectionRate": 0.9122807017543859,
      "gateVerdictCorrectness": 0.7313432835820896,
      "falsePositives": 7,
      "cleanCasesSilent": 6,
      "cleanCases": 13,
      "unscoredCases": 3,
      "casesRun": 70,
      "totalCostUsd": 0.03365288625,
      "meanCostUsdPerReview": 0.00048075551785714283,
      "latencyMsP50": 13429,
      "latencyMsP95": 27855
    },
    {
      "id": "thinkingmachines/inkling-small",
      "detectionRate": 0.8947368421052632,
      "gateVerdictCorrectness": 0.8208955223880597,
      "falsePositives": 0,
      "cleanCasesSilent": 13,
      "cleanCases": 13,
      "unscoredCases": 3,
      "casesRun": 70,
      "totalCostUsd": 0.2467231,
      "meanCostUsdPerReview": 0.0035246157142857144,
      "latencyMsP50": 23097,
      "latencyMsP95": 63238
    },
    {
      "id": "google/gemma-3-27b-it",
      "detectionRate": 0.8771929824561403,
      "gateVerdictCorrectness": 0.7936507936507936,
      "falsePositives": 8,
      "cleanCasesSilent": 4,
      "cleanCases": 13,
      "unscoredCases": 7,
      "casesRun": 70,
      "totalCostUsd": 0.042586394,
      "meanCostUsdPerReview": 0.0006083770571428572,
      "latencyMsP50": 13052,
      "latencyMsP95": 41756
    },
    {
      "id": "deepseek/deepseek-v4-pro-0813",
      "detectionRate": 0.8596491228070176,
      "gateVerdictCorrectness": 0.7857142857142857,
      "falsePositives": 2,
      "cleanCasesSilent": 12,
      "cleanCases": 13,
      "unscoredCases": 0,
      "casesRun": 70,
      "totalCostUsd": 0.70536192,
      "meanCostUsdPerReview": 0.010076598857142858,
      "latencyMsP50": 17953,
      "latencyMsP95": 48373
    },
    {
      "id": "qwen/qwen3-32b",
      "detectionRate": 0.8596491228070176,
      "gateVerdictCorrectness": 0.7777777777777778,
      "falsePositives": 12,
      "cleanCasesSilent": 8,
      "cleanCases": 13,
      "unscoredCases": 7,
      "casesRun": 70,
      "totalCostUsd": 0.13161951,
      "meanCostUsdPerReview": 0.0018802787142857141,
      "latencyMsP50": 5123,
      "latencyMsP95": 19610
    },
    {
      "id": "meta/muse-glimmer-30b",
      "detectionRate": 0.8421052631578947,
      "gateVerdictCorrectness": 0.8285714285714286,
      "falsePositives": 1,
      "cleanCasesSilent": 13,
      "cleanCases": 13,
      "unscoredCases": 0,
      "casesRun": 70,
      "totalCostUsd": 0.28107195,
      "meanCostUsdPerReview": 0.0040153135714285715,
      "latencyMsP50": 24363,
      "latencyMsP95": 54546
    },
    {
      "id": "nvidia/nemotron-3.5-lightning",
      "detectionRate": 0.8421052631578947,
      "gateVerdictCorrectness": 0.7941176470588235,
      "falsePositives": 0,
      "cleanCasesSilent": 13,
      "cleanCases": 13,
      "unscoredCases": 2,
      "casesRun": 70,
      "totalCostUsd": 0.1732027,
      "meanCostUsdPerReview": 0.0024743242857142855,
      "latencyMsP50": 36385,
      "latencyMsP95": 87095
    },
    {
      "id": "google/gemini-3.7-flash",
      "detectionRate": 0.7543859649122807,
      "gateVerdictCorrectness": 0.803030303030303,
      "falsePositives": 0,
      "cleanCasesSilent": 13,
      "cleanCases": 13,
      "unscoredCases": 4,
      "casesRun": 70,
      "totalCostUsd": 0.2323275,
      "meanCostUsdPerReview": 0.0033189642857142858,
      "latencyMsP50": 11189,
      "latencyMsP95": 22047
    },
    {
      "id": "bytedance-seed/seed-2.0-code",
      "detectionRate": 0.7368421052631579,
      "gateVerdictCorrectness": 0.8333333333333334,
      "falsePositives": 2,
      "cleanCasesSilent": 10,
      "cleanCases": 13,
      "unscoredCases": 16,
      "casesRun": 70,
      "totalCostUsd": 0.6598235,
      "meanCostUsdPerReview": 0.00942605,
      "latencyMsP50": 55973,
      "latencyMsP95": 131605
    },
    {
      "id": "deepseek/deepseek-v4-flash-0731",
      "detectionRate": 0.6842105263157895,
      "gateVerdictCorrectness": 0.6323529411764706,
      "falsePositives": 0,
      "cleanCasesSilent": 13,
      "cleanCases": 13,
      "unscoredCases": 2,
      "casesRun": 70,
      "totalCostUsd": 0.064687309,
      "meanCostUsdPerReview": 0.0009241044142857143,
      "latencyMsP50": 24765,
      "latencyMsP95": 74376
    },
    {
      "id": "z-ai/glm-5.3",
      "detectionRate": 0.631578947368421,
      "gateVerdictCorrectness": 0.75,
      "falsePositives": 0,
      "cleanCasesSilent": 12,
      "cleanCases": 13,
      "unscoredCases": 18,
      "casesRun": 70,
      "totalCostUsd": 1.23569172,
      "meanCostUsdPerReview": 0.017652738857142858,
      "latencyMsP50": 69856,
      "latencyMsP95": 134602
    },
    {
      "id": "qwen/qwen3.8-max",
      "unreachable": true,
      "casesRun": 70,
      "unscoredCases": 70
    }
  ],
  "note": "Rows are screening runs on unconstrained provider routing. Where a model carries pinnedProviderContract, that block is the median of runs made under the hosted contract: one pinned upstream provider, zero-retention required, price ceiling applied.",
  "repeatRuns": {
    "note": "Detection rate from repeated screening runs of one unchanged model against one unchanged corpus, on unconstrained provider routing. A run listed in degradedRunIndexes produced no valid envelope on a large share of its cases, so it records an operational failure rather than the model's judgement; it is retained because dropping bad runs is how a benchmark flatters itself, and excluded from the spread, which is a claim about judgement.",
    "models": [
      {
        "id": "openai/gpt-5.6-luna",
        "detectionRates": [
          0.9649122807017544,
          0.9122807017543859,
          0.9298245614035088,
          0.7543859649122807
        ],
        "degradedRunIndexes": [
          3
        ]
      },
      {
        "id": "moonshotai/kimi-k2.7-code",
        "detectionRates": [
          0.9473684210526315,
          0.8947368421052632,
          0.9824561403508771,
          0.9473684210526315
        ]
      },
      {
        "id": "z-ai/glm-5.2",
        "detectionRates": [
          0.9473684210526315,
          0.8947368421052632,
          0.8771929824561403,
          0.8947368421052632
        ]
      }
    ]
  }
}