{
  "metadata": {
    "benchmark": "CUA-bench",
    "slug": "cua_bench",
    "description": "Can an AI agent play six commercial video games with only a keyboard and a mouse? Half of them are held out.",
    "benchmark_id": "cua_bench",
    "family": "cua_bench",
    "version": "1",
    "updated": "2026-10-13",
    "dataset_type": "private",
    "industry": "beta",
    "total_models": 9,
    "models": [
      "anthropic/claude-fable-5-1",
      "anthropic/claude-opus-5",
      "anthropic/claude-opus-5-5",
      "google/gemini-3.8-flash",
      "google/gemini-4-argon",
      "meta/muse-wonder",
      "meta/muse_spark_1_3_max",
      "openai/gpt-5.6-sol",
      "openai/gpt-6-astra"
    ],
    "visible": true,
    "showBadge": false,
    "show_stderr": false,
    "latency_label": "Time / action",
    "partners": [],
    "use_cost_per_test": true,
    "cost_label": "Est. API cost",
    "runner": "external",
    "mode": "agentic",
    "archived": false,
    "partner": false,
    "tasks": {
      "overall": "Overall",
      "minecraft_nether": "Minecraft",
      "held_out_sandbox": "Hidden sandbox",
      "superhot": "SUPERHOT",
      "held_out_reflex": "Hidden reflex",
      "efootball": "eFootball",
      "held_out_sports": "Hidden sports"
    }
  },
  "tasks": {
    "overall": {
      "openai/gpt-6-astra": {
        "accuracy": 19.166667,
        "latency": 32.09,
        "stderr": 0,
        "cost_per_test": 1821.373046,
        "temperature": null,
        "top_p": null,
        "max_output_tokens": 128000,
        "reasoning": true,
        "reasoning_effort": "max",
        "verbosity": null,
        "compute_effort": "max",
        "provider": "openai",
        "harness": "codex",
        "cost_note": "Launch-day runs barely reused the provider's prompt cache, so most of their context was counted as cache writes. A later run with the same setup cached normally."
      },
      "anthropic/claude-fable-5-1": {
        "accuracy": 13.166667,
        "latency": 37.44,
        "stderr": 0,
        "cost_per_test": 323.776608,
        "temperature": null,
        "top_p": null,
        "max_output_tokens": 128000,
        "reasoning": true,
        "reasoning_effort": "max",
        "verbosity": null,
        "compute_effort": "max",
        "provider": "anthropic",
        "harness": "claude-code"
      },
      "anthropic/claude-opus-5": {
        "accuracy": 9.0,
        "latency": 25.29,
        "stderr": 0,
        "cost_per_test": 232.233536,
        "temperature": null,
        "top_p": null,
        "max_output_tokens": 128000,
        "reasoning": true,
        "reasoning_effort": "max",
        "verbosity": null,
        "compute_effort": "max",
        "provider": "anthropic",
        "harness": "claude-code"
      },
      "openai/gpt-5.6-sol": {
        "accuracy": 8.333333,
        "latency": 18.8,
        "stderr": 0,
        "cost_per_test": 206.991579,
        "temperature": null,
        "top_p": null,
        "max_output_tokens": 128000,
        "reasoning": true,
        "reasoning_effort": "max",
        "verbosity": null,
        "compute_effort": "max",
        "provider": "openai",
        "harness": "codex"
      },
      "google/gemini-3.8-flash": {
        "accuracy": 4.166667,
        "latency": 36.55,
        "stderr": 0,
        "cost_per_test": 280.032879,
        "temperature": null,
        "top_p": null,
        "max_output_tokens": 128000,
        "reasoning": true,
        "reasoning_effort": "high",
        "verbosity": null,
        "compute_effort": "high",
        "provider": "google",
        "harness": "google-computer-use"
      },
      "meta/muse_spark_1_3_max": {
        "accuracy": 5.833333,
        "latency": 36.39,
        "stderr": 0,
        "cost_per_test": 143.481477,
        "temperature": null,
        "top_p": null,
        "max_output_tokens": 128000,
        "reasoning": true,
        "reasoning_effort": "max",
        "verbosity": null,
        "compute_effort": "max",
        "provider": "meta",
        "harness": "muse-code"
      },
      "anthropic/claude-opus-5-5": {
        "accuracy": 14.0,
        "latency": 14.71,
        "stderr": 0,
        "cost_per_test": 188.040183,
        "temperature": null,
        "top_p": null,
        "max_output_tokens": 128000,
        "reasoning": true,
        "reasoning_effort": "max",
        "verbosity": null,
        "compute_effort": "max",
        "provider": "anthropic",
        "harness": "claude-code"
      },
      "google/gemini-4-argon": {
        "accuracy": 4.833333,
        "latency": 70.51,
        "stderr": 0,
        "cost_per_test": 193.778921,
        "temperature": null,
        "top_p": null,
        "max_output_tokens": 128000,
        "reasoning": true,
        "reasoning_effort": "high",
        "verbosity": null,
        "compute_effort": "high",
        "provider": "google",
        "harness": "antigravity"
      },
      "meta/muse-wonder": {
        "accuracy": 7.166667,
        "latency": 30.56,
        "stderr": 0,
        "cost_per_test": 597.211635,
        "temperature": null,
        "top_p": null,
        "max_output_tokens": 128000,
        "reasoning": true,
        "reasoning_effort": "max",
        "verbosity": null,
        "compute_effort": "max",
        "provider": "meta",
        "harness": "muse-code"
      }
    },
    "minecraft_nether": {
      "openai/gpt-6-astra": {
        "accuracy": 30,
        "latency": 42.02,
        "stderr": 0,
        "cost_per_test": 162.301085,
        "temperature": null,
        "top_p": null,
        "max_output_tokens": 128000,
        "reasoning": true,
        "reasoning_effort": "max",
        "verbosity": null,
        "compute_effort": "max",
        "provider": "openai",
        "harness": "codex",
        "cost_note": "Launch-day runs barely reused the provider's prompt cache, so most of their context was counted as cache writes. A later run with the same setup cached normally."
      },
      "anthropic/claude-fable-5-1": {
        "accuracy": 25,
        "latency": 51.37,
        "stderr": 0,
        "cost_per_test": 33.945772,
        "temperature": null,
        "top_p": null,
        "max_output_tokens": 128000,
        "reasoning": true,
        "reasoning_effort": "max",
        "verbosity": null,
        "compute_effort": "max",
        "provider": "anthropic",
        "harness": "claude-code"
      },
      "anthropic/claude-opus-5": {
        "accuracy": 30,
        "latency": 35.95,
        "stderr": 0,
        "cost_per_test": 32.804275,
        "temperature": null,
        "top_p": null,
        "max_output_tokens": 128000,
        "reasoning": true,
        "reasoning_effort": "max",
        "verbosity": null,
        "compute_effort": "max",
        "provider": "anthropic",
        "harness": "claude-code"
      },
      "openai/gpt-5.6-sol": {
        "accuracy": 20,
        "latency": 14.61,
        "stderr": 0,
        "cost_per_test": 40.181992,
        "temperature": null,
        "top_p": null,
        "max_output_tokens": 128000,
        "reasoning": true,
        "reasoning_effort": "max",
        "verbosity": null,
        "compute_effort": "max",
        "provider": "openai",
        "harness": "codex"
      },
      "google/gemini-3.8-flash": {
        "accuracy": 5,
        "latency": 40.28,
        "stderr": 0,
        "cost_per_test": 51.830347,
        "temperature": null,
        "top_p": null,
        "max_output_tokens": 128000,
        "reasoning": true,
        "reasoning_effort": "high",
        "verbosity": null,
        "compute_effort": "high",
        "provider": "google",
        "harness": "google-computer-use"
      },
      "meta/muse_spark_1_3_max": {
        "accuracy": 15,
        "latency": 35.83,
        "stderr": 0,
        "cost_per_test": 30.157452,
        "temperature": null,
        "top_p": null,
        "max_output_tokens": 128000,
        "reasoning": true,
        "reasoning_effort": "max",
        "verbosity": null,
        "compute_effort": "max",
        "provider": "meta",
        "harness": "muse-code"
      },
      "anthropic/claude-opus-5-5": {
        "accuracy": 30,
        "latency": 14.56,
        "stderr": 0,
        "cost_per_test": 31.080317,
        "temperature": null,
        "top_p": null,
        "max_output_tokens": 128000,
        "reasoning": true,
        "reasoning_effort": "max",
        "verbosity": null,
        "compute_effort": "max",
        "provider": "anthropic",
        "harness": "claude-code"
      },
      "google/gemini-4-argon": {
        "accuracy": 15,
        "latency": 77.14,
        "stderr": 0,
        "cost_per_test": 19.353702,
        "temperature": null,
        "top_p": null,
        "max_output_tokens": 128000,
        "reasoning": true,
        "reasoning_effort": "high",
        "verbosity": null,
        "compute_effort": "high",
        "provider": "google",
        "harness": "antigravity"
      },
      "meta/muse-wonder": {
        "accuracy": 15,
        "latency": 29.17,
        "stderr": 0,
        "cost_per_test": 110.456332,
        "temperature": null,
        "top_p": null,
        "max_output_tokens": 128000,
        "reasoning": true,
        "reasoning_effort": "max",
        "verbosity": null,
        "compute_effort": "max",
        "provider": "meta",
        "harness": "muse-code"
      }
    },
    "held_out_sandbox": {
      "openai/gpt-6-astra": {
        "accuracy": 15,
        "latency": 30.86,
        "stderr": 0,
        "cost_per_test": 454.590025,
        "temperature": null,
        "top_p": null,
        "max_output_tokens": 128000,
        "reasoning": true,
        "reasoning_effort": "max",
        "verbosity": null,
        "compute_effort": "max",
        "provider": "openai",
        "harness": "codex",
        "cost_note": "Launch-day runs barely reused the provider's prompt cache, so most of their context was counted as cache writes. A later run with the same setup cached normally."
      },
      "anthropic/claude-fable-5-1": {
        "accuracy": 0,
        "latency": 59.13,
        "stderr": 0,
        "cost_per_test": 27.128099,
        "temperature": null,
        "top_p": null,
        "max_output_tokens": 128000,
        "reasoning": true,
        "reasoning_effort": "max",
        "verbosity": null,
        "compute_effort": "max",
        "provider": "anthropic",
        "harness": "claude-code"
      },
      "anthropic/claude-opus-5": {
        "accuracy": 0,
        "latency": 18.03,
        "stderr": 0,
        "cost_per_test": 47.881221,
        "temperature": null,
        "top_p": null,
        "max_output_tokens": 128000,
        "reasoning": true,
        "reasoning_effort": "max",
        "verbosity": null,
        "compute_effort": "max",
        "provider": "anthropic",
        "harness": "claude-code"
      },
      "openai/gpt-5.6-sol": {
        "accuracy": 0,
        "latency": 20.77,
        "stderr": 0,
        "cost_per_test": 32.13623,
        "temperature": null,
        "top_p": null,
        "max_output_tokens": 128000,
        "reasoning": true,
        "reasoning_effort": "max",
        "verbosity": null,
        "compute_effort": "max",
        "provider": "openai",
        "harness": "codex"
      },
      "google/gemini-3.8-flash": {
        "accuracy": 0,
        "latency": 41.62,
        "stderr": 0,
        "cost_per_test": 48.391569,
        "temperature": null,
        "top_p": null,
        "max_output_tokens": 128000,
        "reasoning": true,
        "reasoning_effort": "high",
        "verbosity": null,
        "compute_effort": "high",
        "provider": "google",
        "harness": "google-computer-use"
      },
      "meta/muse_spark_1_3_max": {
        "accuracy": 0,
        "latency": 37.76,
        "stderr": 0,
        "cost_per_test": 26.772657,
        "temperature": null,
        "top_p": null,
        "max_output_tokens": 128000,
        "reasoning": true,
        "reasoning_effort": "max",
        "verbosity": null,
        "compute_effort": "max",
        "provider": "meta",
        "harness": "muse-code"
      },
      "anthropic/claude-opus-5-5": {
        "accuracy": 20,
        "latency": 15.16,
        "stderr": 0,
        "cost_per_test": 28.012298,
        "temperature": null,
        "top_p": null,
        "max_output_tokens": 128000,
        "reasoning": true,
        "reasoning_effort": "max",
        "verbosity": null,
        "compute_effort": "max",
        "provider": "anthropic",
        "harness": "claude-code"
      },
      "google/gemini-4-argon": {
        "accuracy": 0,
        "latency": 68.35,
        "stderr": 0,
        "cost_per_test": 20.086706,
        "temperature": null,
        "top_p": null,
        "max_output_tokens": 128000,
        "reasoning": true,
        "reasoning_effort": "high",
        "verbosity": null,
        "compute_effort": "high",
        "provider": "google",
        "harness": "antigravity"
      },
      "meta/muse-wonder": {
        "accuracy": 0,
        "latency": 34.95,
        "stderr": 0,
        "cost_per_test": 86.595802,
        "temperature": null,
        "top_p": null,
        "max_output_tokens": 128000,
        "reasoning": true,
        "reasoning_effort": "max",
        "verbosity": null,
        "compute_effort": "max",
        "provider": "meta",
        "harness": "muse-code"
      }
    },
    "superhot": {
      "openai/gpt-6-astra": {
        "accuracy": 20,
        "latency": 24.16,
        "stderr": 0,
        "cost_per_test": 593.228319,
        "temperature": null,
        "top_p": null,
        "max_output_tokens": 128000,
        "reasoning": true,
        "reasoning_effort": "max",
        "verbosity": null,
        "compute_effort": "max",
        "provider": "openai",
        "harness": "codex",
        "cost_note": "Launch-day runs barely reused the provider's prompt cache, so most of their context was counted as cache writes. A later run with the same setup cached normally."
      },
      "anthropic/claude-fable-5-1": {
        "accuracy": 12,
        "latency": 30.68,
        "stderr": 0,
        "cost_per_test": 98.90788,
        "temperature": null,
        "top_p": null,
        "max_output_tokens": 128000,
        "reasoning": true,
        "reasoning_effort": "max",
        "verbosity": null,
        "compute_effort": "max",
        "provider": "anthropic",
        "harness": "claude-code"
      },
      "anthropic/claude-opus-5": {
        "accuracy": 4,
        "latency": 20.3,
        "stderr": 0,
        "cost_per_test": 49.093325,
        "temperature": null,
        "top_p": null,
        "max_output_tokens": 128000,
        "reasoning": true,
        "reasoning_effort": "max",
        "verbosity": null,
        "compute_effort": "max",
        "provider": "anthropic",
        "harness": "claude-code"
      },
      "openai/gpt-5.6-sol": {
        "accuracy": 8,
        "latency": 18.43,
        "stderr": 0,
        "cost_per_test": 36.704619,
        "temperature": null,
        "top_p": null,
        "max_output_tokens": 128000,
        "reasoning": true,
        "reasoning_effort": "max",
        "verbosity": null,
        "compute_effort": "max",
        "provider": "openai",
        "harness": "codex"
      },
      "google/gemini-3.8-flash": {
        "accuracy": 0,
        "latency": 34.29,
        "stderr": 0,
        "cost_per_test": 51.308013,
        "temperature": null,
        "top_p": null,
        "max_output_tokens": 128000,
        "reasoning": true,
        "reasoning_effort": "high",
        "verbosity": null,
        "compute_effort": "high",
        "provider": "google",
        "harness": "google-computer-use"
      },
      "meta/muse_spark_1_3_max": {
        "accuracy": 0,
        "latency": 38.55,
        "stderr": 0,
        "cost_per_test": 29.344737,
        "temperature": null,
        "top_p": null,
        "max_output_tokens": 128000,
        "reasoning": true,
        "reasoning_effort": "max",
        "verbosity": null,
        "compute_effort": "max",
        "provider": "meta",
        "harness": "muse-code"
      },
      "anthropic/claude-opus-5-5": {
        "accuracy": 12,
        "latency": 15.95,
        "stderr": 0,
        "cost_per_test": 29.393261,
        "temperature": null,
        "top_p": null,
        "max_output_tokens": 128000,
        "reasoning": true,
        "reasoning_effort": "max",
        "verbosity": null,
        "compute_effort": "max",
        "provider": "anthropic",
        "harness": "claude-code"
      },
      "google/gemini-4-argon": {
        "accuracy": 4,
        "latency": 103.85,
        "stderr": 0,
        "cost_per_test": 50.113634,
        "temperature": null,
        "top_p": null,
        "max_output_tokens": 128000,
        "reasoning": true,
        "reasoning_effort": "high",
        "verbosity": null,
        "compute_effort": "high",
        "provider": "google",
        "harness": "antigravity"
      },
      "meta/muse-wonder": {
        "accuracy": 8,
        "latency": 31.5,
        "stderr": 0,
        "cost_per_test": 99.046993,
        "temperature": null,
        "top_p": null,
        "max_output_tokens": 128000,
        "reasoning": true,
        "reasoning_effort": "max",
        "verbosity": null,
        "compute_effort": "max",
        "provider": "meta",
        "harness": "muse-code"
      }
    },
    "held_out_reflex": {
      "openai/gpt-6-astra": {
        "accuracy": 20,
        "latency": 44.08,
        "stderr": 0,
        "cost_per_test": 215.849862,
        "temperature": null,
        "top_p": null,
        "max_output_tokens": 128000,
        "reasoning": true,
        "reasoning_effort": "max",
        "verbosity": null,
        "compute_effort": "max",
        "provider": "openai",
        "harness": "codex",
        "cost_note": "Launch-day runs barely reused the provider's prompt cache, so most of their context was counted as cache writes. A later run with the same setup cached normally."
      },
      "anthropic/claude-fable-5-1": {
        "accuracy": 12,
        "latency": 38.13,
        "stderr": 0,
        "cost_per_test": 37.693043,
        "temperature": null,
        "top_p": null,
        "max_output_tokens": 128000,
        "reasoning": true,
        "reasoning_effort": "max",
        "verbosity": null,
        "compute_effort": "max",
        "provider": "anthropic",
        "harness": "claude-code"
      },
      "anthropic/claude-opus-5": {
        "accuracy": 20,
        "latency": 36.47,
        "stderr": 0,
        "cost_per_test": 31.458462,
        "temperature": null,
        "top_p": null,
        "max_output_tokens": 128000,
        "reasoning": true,
        "reasoning_effort": "max",
        "verbosity": null,
        "compute_effort": "max",
        "provider": "anthropic",
        "harness": "claude-code"
      },
      "openai/gpt-5.6-sol": {
        "accuracy": 22,
        "latency": 15.02,
        "stderr": 0,
        "cost_per_test": 43.544694,
        "temperature": null,
        "top_p": null,
        "max_output_tokens": 128000,
        "reasoning": true,
        "reasoning_effort": "max",
        "verbosity": null,
        "compute_effort": "max",
        "provider": "openai",
        "harness": "codex"
      },
      "google/gemini-3.8-flash": {
        "accuracy": 20,
        "latency": 37.71,
        "stderr": 0,
        "cost_per_test": 41.807782,
        "temperature": null,
        "top_p": null,
        "max_output_tokens": 128000,
        "reasoning": true,
        "reasoning_effort": "high",
        "verbosity": null,
        "compute_effort": "high",
        "provider": "google",
        "harness": "google-computer-use"
      },
      "meta/muse_spark_1_3_max": {
        "accuracy": 20,
        "latency": 36.09,
        "stderr": 0,
        "cost_per_test": 25.076925,
        "temperature": null,
        "top_p": null,
        "max_output_tokens": 128000,
        "reasoning": true,
        "reasoning_effort": "max",
        "verbosity": null,
        "compute_effort": "max",
        "provider": "meta",
        "harness": "muse-code"
      },
      "anthropic/claude-opus-5-5": {
        "accuracy": 22,
        "latency": 18.6,
        "stderr": 0,
        "cost_per_test": 25.365201,
        "temperature": null,
        "top_p": null,
        "max_output_tokens": 128000,
        "reasoning": true,
        "reasoning_effort": "max",
        "verbosity": null,
        "compute_effort": "max",
        "provider": "anthropic",
        "harness": "claude-code"
      },
      "google/gemini-4-argon": {
        "accuracy": 10,
        "latency": 80.6,
        "stderr": 0,
        "cost_per_test": 28.818046,
        "temperature": null,
        "top_p": null,
        "max_output_tokens": 128000,
        "reasoning": true,
        "reasoning_effort": "high",
        "verbosity": null,
        "compute_effort": "high",
        "provider": "google",
        "harness": "antigravity"
      },
      "meta/muse-wonder": {
        "accuracy": 20,
        "latency": 32.7,
        "stderr": 0,
        "cost_per_test": 109.966906,
        "temperature": null,
        "top_p": null,
        "max_output_tokens": 128000,
        "reasoning": true,
        "reasoning_effort": "max",
        "verbosity": null,
        "compute_effort": "max",
        "provider": "meta",
        "harness": "muse-code"
      }
    },
    "efootball": {
      "openai/gpt-6-astra": {
        "accuracy": 30,
        "latency": 27.69,
        "stderr": 0,
        "cost_per_test": 341.382408,
        "temperature": null,
        "top_p": null,
        "max_output_tokens": 128000,
        "reasoning": true,
        "reasoning_effort": "max",
        "verbosity": null,
        "compute_effort": "max",
        "provider": "openai",
        "harness": "codex",
        "cost_note": "Launch-day runs barely reused the provider's prompt cache, so most of their context was counted as cache writes. A later run with the same setup cached normally."
      },
      "anthropic/claude-fable-5-1": {
        "accuracy": 30,
        "latency": 22.95,
        "stderr": 0,
        "cost_per_test": 87.02114,
        "temperature": null,
        "top_p": null,
        "max_output_tokens": 128000,
        "reasoning": true,
        "reasoning_effort": "max",
        "verbosity": null,
        "compute_effort": "max",
        "provider": "anthropic",
        "harness": "claude-code"
      },
      "anthropic/claude-opus-5": {
        "accuracy": 0,
        "latency": 18.66,
        "stderr": 0,
        "cost_per_test": 44.39607,
        "temperature": null,
        "top_p": null,
        "max_output_tokens": 128000,
        "reasoning": true,
        "reasoning_effort": "max",
        "verbosity": null,
        "compute_effort": "max",
        "provider": "anthropic",
        "harness": "claude-code"
      },
      "openai/gpt-5.6-sol": {
        "accuracy": 0,
        "latency": 22.45,
        "stderr": 0,
        "cost_per_test": 29.607119,
        "temperature": null,
        "top_p": null,
        "max_output_tokens": 128000,
        "reasoning": true,
        "reasoning_effort": "max",
        "verbosity": null,
        "compute_effort": "max",
        "provider": "openai",
        "harness": "codex"
      },
      "google/gemini-3.8-flash": {
        "accuracy": 0,
        "latency": 34.4,
        "stderr": 0,
        "cost_per_test": 67.980694,
        "temperature": null,
        "top_p": null,
        "max_output_tokens": 128000,
        "reasoning": true,
        "reasoning_effort": "high",
        "verbosity": null,
        "compute_effort": "high",
        "provider": "google",
        "harness": "google-computer-use"
      },
      "meta/muse_spark_1_3_max": {
        "accuracy": 0,
        "latency": 35.49,
        "stderr": 0,
        "cost_per_test": 27.37126,
        "temperature": null,
        "top_p": null,
        "max_output_tokens": 128000,
        "reasoning": true,
        "reasoning_effort": "max",
        "verbosity": null,
        "compute_effort": "max",
        "provider": "meta",
        "harness": "muse-code"
      },
      "anthropic/claude-opus-5-5": {
        "accuracy": 0,
        "latency": 12.94,
        "stderr": 0,
        "cost_per_test": 39.906343,
        "temperature": null,
        "top_p": null,
        "max_output_tokens": 128000,
        "reasoning": true,
        "reasoning_effort": "max",
        "verbosity": null,
        "compute_effort": "max",
        "provider": "anthropic",
        "harness": "claude-code"
      },
      "google/gemini-4-argon": {
        "accuracy": 0,
        "latency": 117.39,
        "stderr": 0,
        "cost_per_test": 17.228613,
        "temperature": null,
        "top_p": null,
        "max_output_tokens": 128000,
        "reasoning": true,
        "reasoning_effort": "high",
        "verbosity": null,
        "compute_effort": "high",
        "provider": "google",
        "harness": "antigravity"
      },
      "meta/muse-wonder": {
        "accuracy": 0,
        "latency": 27.28,
        "stderr": 0,
        "cost_per_test": 92.471983,
        "temperature": null,
        "top_p": null,
        "max_output_tokens": 128000,
        "reasoning": true,
        "reasoning_effort": "max",
        "verbosity": null,
        "compute_effort": "max",
        "provider": "meta",
        "harness": "muse-code"
      }
    },
    "held_out_sports": {
      "openai/gpt-6-astra": {
        "accuracy": 0,
        "latency": 32.74,
        "stderr": 0,
        "cost_per_test": 54.021346,
        "temperature": null,
        "top_p": null,
        "max_output_tokens": 128000,
        "reasoning": true,
        "reasoning_effort": "max",
        "verbosity": null,
        "compute_effort": "max",
        "provider": "openai",
        "harness": "codex"
      },
      "anthropic/claude-fable-5-1": {
        "accuracy": 0,
        "latency": 47.3,
        "stderr": 0,
        "cost_per_test": 39.080675,
        "temperature": null,
        "top_p": null,
        "max_output_tokens": 128000,
        "reasoning": true,
        "reasoning_effort": "max",
        "verbosity": null,
        "compute_effort": "max",
        "provider": "anthropic",
        "harness": "claude-code"
      },
      "anthropic/claude-opus-5": {
        "accuracy": 0,
        "latency": 42.38,
        "stderr": 0,
        "cost_per_test": 26.600182,
        "temperature": null,
        "top_p": null,
        "max_output_tokens": 128000,
        "reasoning": true,
        "reasoning_effort": "max",
        "verbosity": null,
        "compute_effort": "max",
        "provider": "anthropic",
        "harness": "claude-code"
      },
      "openai/gpt-5.6-sol": {
        "accuracy": 0,
        "latency": 26.93,
        "stderr": 0,
        "cost_per_test": 24.816925,
        "temperature": null,
        "top_p": null,
        "max_output_tokens": 128000,
        "reasoning": true,
        "reasoning_effort": "max",
        "verbosity": null,
        "compute_effort": "max",
        "provider": "openai",
        "harness": "codex"
      },
      "google/gemini-3.8-flash": {
        "accuracy": 0,
        "latency": 29.55,
        "stderr": 0,
        "cost_per_test": 18.714475,
        "temperature": null,
        "top_p": null,
        "max_output_tokens": 128000,
        "reasoning": true,
        "reasoning_effort": "high",
        "verbosity": null,
        "compute_effort": "high",
        "provider": "google",
        "harness": "google-computer-use"
      },
      "meta/muse_spark_1_3_max": {
        "accuracy": 0,
        "latency": 30.67,
        "stderr": 0,
        "cost_per_test": 4.758446,
        "temperature": null,
        "top_p": null,
        "max_output_tokens": 128000,
        "reasoning": true,
        "reasoning_effort": "max",
        "verbosity": null,
        "compute_effort": "max",
        "provider": "meta",
        "harness": "muse-code"
      },
      "anthropic/claude-opus-5-5": {
        "accuracy": 0,
        "latency": 12.54,
        "stderr": 0,
        "cost_per_test": 34.282763,
        "temperature": null,
        "top_p": null,
        "max_output_tokens": 128000,
        "reasoning": true,
        "reasoning_effort": "max",
        "verbosity": null,
        "compute_effort": "max",
        "provider": "anthropic",
        "harness": "claude-code"
      },
      "google/gemini-4-argon": {
        "accuracy": 0,
        "latency": 37.11,
        "stderr": 0,
        "cost_per_test": 58.178219,
        "temperature": null,
        "top_p": null,
        "max_output_tokens": 128000,
        "reasoning": true,
        "reasoning_effort": "high",
        "verbosity": null,
        "compute_effort": "high",
        "provider": "google",
        "harness": "antigravity"
      },
      "meta/muse-wonder": {
        "accuracy": 0,
        "latency": 28.92,
        "stderr": 0,
        "cost_per_test": 98.673619,
        "temperature": null,
        "top_p": null,
        "max_output_tokens": 128000,
        "reasoning": true,
        "reasoning_effort": "max",
        "verbosity": null,
        "compute_effort": "max",
        "provider": "meta",
        "harness": "muse-code"
      }
    }
  }
}