Independent benchmark · harness history

Claude Code

On X: @claudeai
Security leaderboard · Claude Code

Every recorded run

The latest run shows its results, with every earlier published version below it. A disclosure is shown against the run that raised it; a prototype run appears only if it raised a disclosure, by its date, with its results ghosted out.

RunTestedFailsGapsS01S02S03S04S05S06S07S09S10Disclosure
2.1.308latest run2026-10-222026-10-2210✓✓✓✓✗✓✓✓✓W
2.1.286earlier version2026-10-172026-10-1710✓✓✓✓✗✓✓✓✓W
Prototype run2026-09-262026-09-26W

Back to the leaderboard