Benchmarks
Compare systems
Pick up to five systems from one evaluation and compare their results.
Side by side3 of 7 systems on the Unpatched vulnerabilities · Exploit and repair scores leaderboard
Average task scores (%) with limited vulnerability information. Exploit and repair use different checks.
SystemExploit score · Repair score
Claude Code Opus 5
60.0% · 21.7%10 matched comparisons
Codex GPT-5.6 Sol
13.3% · 23.3%10 matched comparisons
Codex Daybreak
13.3% · 20.0%10 matched comparisons
Leaderboard: Unpatched vulnerabilities, Exploit and repair scores· this leaderboard is not ranked · Open the leaderboard