{
  "schema_version": 1,
  "method": "Unweighted mean of two original per-observation rates; exact rational arithmetic.",
  "cost_assumptions": "River uses owner-selected $0.12/$0.01/$0.50 per million input/cached-input/output tokens; ranges compare hypothetical all-cached versus no-cache usage. Other estimates retain campaign tariffs.",
  "models": [
    {
      "model": "gpt-6-astra",
      "average_quality_percent": "96.6%",
      "average_goal_achieved_percent": "100.0%",
      "quality_rate": {
        "numerator": 1827,
        "denominator": 1892
      },
      "goal_rate": {
        "numerator": 1,
        "denominator": 1
      },
      "average_full_run_estimate": {
        "model": "gpt-6-astra",
        "mean_cost_low_usd": "238.130875",
        "mean_cost_high_usd": "238.130875",
        "incomplete_usage": true,
        "pricing_basis": "retained_tariff_estimate"
      }
    },
    {
      "model": "gemini-4-argon",
      "average_quality_percent": "95.5%",
      "average_goal_achieved_percent": "100.0%",
      "quality_rate": {
        "numerator": 21,
        "denominator": 22
      },
      "goal_rate": {
        "numerator": 1,
        "denominator": 1
      },
      "average_full_run_estimate": {
        "model": "gemini-4-argon",
        "mean_cost_low_usd": "94.566744",
        "mean_cost_high_usd": "94.566744",
        "incomplete_usage": false,
        "pricing_basis": "retained_tariff_estimate"
      }
    },
    {
      "model": "gpt-6.1-sol",
      "average_quality_percent": "95.5%",
      "average_goal_achieved_percent": "100.0%",
      "quality_rate": {
        "numerator": 21,
        "denominator": 22
      },
      "goal_rate": {
        "numerator": 1,
        "denominator": 1
      },
      "average_full_run_estimate": {
        "model": "gpt-6.1-sol",
        "mean_cost_low_usd": "52.671586",
        "mean_cost_high_usd": "52.671586",
        "incomplete_usage": false,
        "pricing_basis": "retained_tariff_estimate"
      }
    },
    {
      "model": "claude-fable-5-1",
      "average_quality_percent": "80.5%",
      "average_goal_achieved_percent": "100.0%",
      "quality_rate": {
        "numerator": 3045,
        "denominator": 3784
      },
      "goal_rate": {
        "numerator": 1,
        "denominator": 1
      },
      "average_full_run_estimate": {
        "model": "claude-fable-5-1",
        "mean_cost_low_usd": "192.630185",
        "mean_cost_high_usd": "192.630185",
        "incomplete_usage": true,
        "pricing_basis": "retained_tariff_estimate"
      }
    },
    {
      "model": "claude-opus-5-5",
      "average_quality_percent": "54.5%",
      "average_goal_achieved_percent": "98.9%",
      "quality_rate": {
        "numerator": 6,
        "denominator": 11
      },
      "goal_rate": {
        "numerator": 87,
        "denominator": 88
      },
      "average_full_run_estimate": {
        "model": "claude-opus-5-5",
        "mean_cost_low_usd": "153.684212",
        "mean_cost_high_usd": "153.684212",
        "incomplete_usage": false,
        "pricing_basis": "retained_tariff_estimate"
      }
    },
    {
      "model": "gemini-3.8-flash",
      "average_quality_percent": "45.3%",
      "average_goal_achieved_percent": "100.0%",
      "quality_rate": {
        "numerator": 39,
        "denominator": 86
      },
      "goal_rate": {
        "numerator": 1,
        "denominator": 1
      },
      "average_full_run_estimate": {
        "model": "gemini-3.8-flash",
        "mean_cost_low_usd": "29.0916735",
        "mean_cost_high_usd": "29.0916735",
        "incomplete_usage": true,
        "pricing_basis": "retained_tariff_estimate"
      }
    },
    {
      "model": "gpt-6-sol",
      "average_quality_percent": "44.3%",
      "average_goal_achieved_percent": "89.8%",
      "quality_rate": {
        "numerator": 39,
        "denominator": 88
      },
      "goal_rate": {
        "numerator": 79,
        "denominator": 88
      },
      "average_full_run_estimate": {
        "model": "gpt-6-sol",
        "mean_cost_low_usd": "74.553171",
        "mean_cost_high_usd": "74.553171",
        "incomplete_usage": false,
        "pricing_basis": "retained_tariff_estimate"
      }
    },
    {
      "model": "muse-wonder-1.0",
      "average_quality_percent": "26.5%",
      "average_goal_achieved_percent": "56.4%",
      "quality_rate": {
        "numerator": 91,
        "denominator": 344
      },
      "goal_rate": {
        "numerator": 97,
        "denominator": 172
      },
      "average_full_run_estimate": {
        "model": "muse-wonder-1.0",
        "mean_cost_low_usd": "52.067637",
        "mean_cost_high_usd": "52.067637",
        "incomplete_usage": true,
        "pricing_basis": "retained_tariff_estimate"
      }
    },
    {
      "model": "zai-org/GLM-5.3-Flash",
      "average_quality_percent": "14.8%",
      "average_goal_achieved_percent": "77.3%",
      "quality_rate": {
        "numerator": 13,
        "denominator": 88
      },
      "goal_rate": {
        "numerator": 17,
        "denominator": 22
      },
      "average_full_run_estimate": {
        "model": "zai-org/GLM-5.3-Flash",
        "mean_cost_low_usd": "0.18880107",
        "mean_cost_high_usd": "1.20868609",
        "incomplete_usage": false,
        "pricing_basis": "river_owner_cache_scenarios"
      }
    },
    {
      "model": "nvidia/GLM-5.2-NVFP4-262K",
      "average_quality_percent": "13.6%",
      "average_goal_achieved_percent": "71.6%",
      "quality_rate": {
        "numerator": 3,
        "denominator": 22
      },
      "goal_rate": {
        "numerator": 63,
        "denominator": 88
      },
      "average_full_run_estimate": {
        "model": "nvidia/GLM-5.2-NVFP4-262K",
        "mean_cost_low_usd": "0.3451353",
        "mean_cost_high_usd": "1.7250116",
        "incomplete_usage": false,
        "pricing_basis": "river_owner_cache_scenarios"
      }
    },
    {
      "model": "muse-spark-1.3",
      "average_quality_percent": "10.2%",
      "average_goal_achieved_percent": "42.0%",
      "quality_rate": {
        "numerator": 9,
        "denominator": 88
      },
      "goal_rate": {
        "numerator": 37,
        "denominator": 88
      },
      "average_full_run_estimate": {
        "model": "muse-spark-1.3",
        "mean_cost_low_usd": "46.821199625",
        "mean_cost_high_usd": "46.821199625",
        "incomplete_usage": false,
        "pricing_basis": "retained_tariff_estimate"
      }
    },
    {
      "model": "Qwen/Qwen3.8-27B-FP8",
      "average_quality_percent": "3.4%",
      "average_goal_achieved_percent": "3.4%",
      "quality_rate": {
        "numerator": 3,
        "denominator": 88
      },
      "goal_rate": {
        "numerator": 3,
        "denominator": 88
      },
      "average_full_run_estimate": {
        "model": "Qwen/Qwen3.8-27B-FP8",
        "mean_cost_low_usd": "0.29128106",
        "mean_cost_high_usd": "2.52077547",
        "incomplete_usage": true,
        "pricing_basis": "river_owner_cache_scenarios"
      }
    },
    {
      "model": "deepseek-ai/DeepSeek-V4.1-Flash",
      "average_quality_percent": "2.3%",
      "average_goal_achieved_percent": "18.4%",
      "quality_rate": {
        "numerator": 1,
        "denominator": 43
      },
      "goal_rate": {
        "numerator": 695,
        "denominator": 3784
      },
      "average_full_run_estimate": {
        "model": "deepseek-ai/DeepSeek-V4.1-Flash",
        "mean_cost_low_usd": "0.437959705",
        "mean_cost_high_usd": "2.90048796",
        "incomplete_usage": true,
        "pricing_basis": "river_owner_cache_scenarios"
      }
    },
    {
      "model": "deepseek-ai/DeepSeek-V4-Flash-0731",
      "average_quality_percent": "2.3%",
      "average_goal_achieved_percent": "5.7%",
      "quality_rate": {
        "numerator": 1,
        "denominator": 44
      },
      "goal_rate": {
        "numerator": 5,
        "denominator": 88
      },
      "average_full_run_estimate": {
        "model": "deepseek-ai/DeepSeek-V4-Flash-0731",
        "mean_cost_low_usd": "0.3338015",
        "mean_cost_high_usd": "2.54243925",
        "incomplete_usage": true,
        "pricing_basis": "river_owner_cache_scenarios"
      }
    },
    {
      "model": "gpt-6-luna",
      "average_quality_percent": "0.0%",
      "average_goal_achieved_percent": "21.5%",
      "quality_rate": {
        "numerator": 0,
        "denominator": 1
      },
      "goal_rate": {
        "numerator": 189,
        "denominator": 880
      },
      "average_full_run_estimate": {
        "model": "gpt-6-luna",
        "mean_cost_low_usd": "1.42",
        "mean_cost_high_usd": "1.42",
        "incomplete_usage": true,
        "pricing_basis": "retained_tariff_estimate"
      }
    },
    {
      "model": "Qwen/Qwen3.6-35B-A3B-FP8",
      "average_quality_percent": "0.0%",
      "average_goal_achieved_percent": "17.0%",
      "quality_rate": {
        "numerator": 0,
        "denominator": 1
      },
      "goal_rate": {
        "numerator": 15,
        "denominator": 88
      },
      "average_full_run_estimate": {
        "model": "Qwen/Qwen3.6-35B-A3B-FP8",
        "mean_cost_low_usd": "0.419280145",
        "mean_cost_high_usd": "1.95993624",
        "incomplete_usage": false,
        "pricing_basis": "river_owner_cache_scenarios"
      }
    },
    {
      "model": "nvidia/Kimi-K2.6-NVFP4",
      "average_quality_percent": "0.0%",
      "average_goal_achieved_percent": "16.1%",
      "quality_rate": {
        "numerator": 0,
        "denominator": 1
      },
      "goal_rate": {
        "numerator": 609,
        "denominator": 3784
      },
      "average_full_run_estimate": {
        "model": "nvidia/Kimi-K2.6-NVFP4",
        "mean_cost_low_usd": "0.157862835",
        "mean_cost_high_usd": "1.00285627",
        "incomplete_usage": true,
        "pricing_basis": "river_owner_cache_scenarios"
      }
    },
    {
      "model": "Qwen/Qwen3.5-397B-A17B-FP8",
      "average_quality_percent": "0.0%",
      "average_goal_achieved_percent": "6.8%",
      "quality_rate": {
        "numerator": 0,
        "denominator": 1
      },
      "goal_rate": {
        "numerator": 3,
        "denominator": 44
      },
      "average_full_run_estimate": {
        "model": "Qwen/Qwen3.5-397B-A17B-FP8",
        "mean_cost_low_usd": "0.19456902",
        "mean_cost_high_usd": "1.35832249",
        "incomplete_usage": false,
        "pricing_basis": "river_owner_cache_scenarios"
      }
    },
    {
      "model": "mistral-medium-3-5",
      "average_quality_percent": "0.0%",
      "average_goal_achieved_percent": "0.0%",
      "quality_rate": {
        "numerator": 0,
        "denominator": 1
      },
      "goal_rate": {
        "numerator": 0,
        "denominator": 1
      },
      "average_full_run_estimate": {
        "model": "mistral-medium-3-5",
        "mean_cost_low_usd": "25.77",
        "mean_cost_high_usd": "25.77",
        "incomplete_usage": true,
        "pricing_basis": "retained_tariff_estimate"
      }
    },
    {
      "model": "nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-NVFP4",
      "average_quality_percent": "0.0%",
      "average_goal_achieved_percent": "0.0%",
      "quality_rate": {
        "numerator": 0,
        "denominator": 1
      },
      "goal_rate": {
        "numerator": 0,
        "denominator": 1
      },
      "average_full_run_estimate": {
        "model": "nvidia/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-NVFP4",
        "mean_cost_low_usd": "0.2283602",
        "mean_cost_high_usd": "2.2550519",
        "incomplete_usage": false,
        "pricing_basis": "river_owner_cache_scenarios"
      }
    },
    {
      "model": "Qwen/Qwen3.5-9B",
      "average_quality_percent": "0.0%",
      "average_goal_achieved_percent": "0.0%",
      "quality_rate": {
        "numerator": 0,
        "denominator": 1
      },
      "goal_rate": {
        "numerator": 0,
        "denominator": 1
      },
      "average_full_run_estimate": {
        "model": "Qwen/Qwen3.5-9B",
        "mean_cost_low_usd": "0.25020603",
        "mean_cost_high_usd": "1.26602886",
        "incomplete_usage": false,
        "pricing_basis": "river_owner_cache_scenarios"
      }
    }
  ]
}
