Benchmarks
Compare systems
Pick up to five systems from one evaluation and compare their results.
Side by side3 of 3 systems on the Vulnerability repair · Exploit leaderboard
Each bar is the share of vulnerabilities where the system's exploit worked. The reasoning effort for these runs is not recorded.
RankSystemPass rate
1
OpenCode GPT-5.5
96.2%92.3–98.9211 completed runs
2
Claude Code Opus 4.7
79.7%68.6–90.1212 completed runs
2
OpenCode GPT-5.3 Codex
77.3%69.2–84.3211 completed runs
Leaderboard: Vulnerability repair, Exploit· Open the leaderboard