{
  "results": [
    {
      "model": "Claude Opus 5",
      "harness": "AWS Continuum",
      "task_set": 920,
      "budget": "No cost cap / 90 min",
      "patch_only": 93.4,
      "s1": 92.5,
      "s2": 89.6,
      "s3": 89.0,
      "s4": 37.8,
      "source": "AWS Continuum",
      "source_url": "https://aws.amazon.com/blogs/security/aws-continuum-sets-a-new-standard-in-autonomous-code-security/",
      "date": "2026-10-05",
      "focus": "agent"
    },
    {
      "harness": "Aegis",
      "model": "DeepSeek-V4-Flash-0731",
      "task_set": 920,
      "budget": "self-hosted / 90 min",
      "patch_only": 55.7,
      "s1": 32.5,
      "s2": 29.2,
      "s3": 28.4,
      "s4": 10.3,
      "source": "ZJU Aegis",
      "source_url": "https://github.com/bofeng17/Cybergym-E2E/tree/main",
      "date": "2026-09-16",
      "focus": "agent"
    },
    {
      "model": "Claude Opus 4.5",
      "harness": "Claude Code",
      "task_set": 615,
      "budget": "$10 / 90 min",
      "patch_only": 82.3,
      "s1": 24.9,
      "s2": 21.9,
      "s3": 19.2,
      "s4": 7.6,
      "source": "CyberGym-E2E Team",
      "date": "2026-06-03",
      "focus": "model"
    },
    {
      "model": "Claude Sonnet 4.5",
      "harness": "Claude Code",
      "task_set": 615,
      "budget": "$10 / 90 min",
      "patch_only": 77.4,
      "s1": 18.1,
      "s2": 12.1,
      "s3": 10.6,
      "s4": 3.4,
      "source": "CyberGym-E2E Team",
      "date": "2026-06-03",
      "focus": "model"
    },
    {
      "model": "Claude Sonnet 4.5",
      "harness": "OpenHands",
      "task_set": 615,
      "budget": "$10 / 90 min",
      "patch_only": 68.9,
      "s1": 9.3,
      "s2": 7.2,
      "s3": 5.4,
      "s4": 2.3,
      "source": "CyberGym-E2E Team",
      "date": "2026-06-03",
      "focus": "model"
    },
    {
      "model": "GPT-5.2",
      "harness": "Codex",
      "task_set": 615,
      "budget": "$10 / 90 min",
      "patch_only": 58.5,
      "s1": 30.2,
      "s2": 22.0,
      "s3": 20.7,
      "s4": 6.5,
      "source": "CyberGym-E2E Team",
      "date": "2026-06-03",
      "focus": "model"
    },
    {
      "model": "Gemini 3 Pro",
      "harness": "Gemini CLI",
      "task_set": 615,
      "budget": "$10 / 90 min",
      "patch_only": 77.6,
      "s1": 29.6,
      "s2": 23.6,
      "s3": 22.6,
      "s4": 5.0,
      "source": "CyberGym-E2E Team",
      "date": "2026-06-03",
      "focus": "model"
    },
    {
      "model": "Claude Opus 4.6",
      "harness": "Claude Code",
      "task_set": 920,
      "budget": "$10 / 90 min",
      "patch_only": 84.1,
      "s1": 39.7,
      "s2": 39.5,
      "s3": 37.9,
      "s4": 15.7,
      "source": "CyberGym-E2E Team",
      "date": "2026-06-03",
      "focus": "model"
    },
    {
      "model": "GPT-5.4",
      "harness": "Codex",
      "task_set": 920,
      "budget": "$10 / 90 min",
      "patch_only": 87.1,
      "s1": 67.9,
      "s2": 66.2,
      "s3": 65.9,
      "s4": 22.2,
      "source": "CyberGym-E2E Team",
      "date": "2026-06-03",
      "focus": "model"
    },
    {
      "model": "Gemini 3.1 Pro",
      "harness": "Gemini CLI",
      "task_set": 920,
      "budget": "$10 / 90 min",
      "patch_only": 83.0,
      "s1": 47.4,
      "s2": 44.3,
      "s3": 43.8,
      "s4": 20.5,
      "source": "CyberGym-E2E Team",
      "date": "2026-06-03",
      "focus": "model"
    },
    {
      "model": "Claude Opus 4.6",
      "harness": "Claude Code",
      "task_set": 920,
      "budget": "No cost cap / 90 min",
      "patch_only": 85.8,
      "s1": 66.3,
      "s2": 65.0,
      "s3": 62.6,
      "s4": 26.2,
      "source": "CyberGym-E2E Team",
      "date": "2026-06-03",
      "focus": "model"
    }
  ]
}
