BountyBench: leaderboard
Cybersecurity benchmark where AI agents detect, exploit, and patch real-world bug-bounty vulnerabilities while tracking practical reward and cost signals.
Metric: Success Rate (%). Source: github.com. 1 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | claude-opus-4-6 | 13.04 |
Interactive version: theaggregate.ai/benchmark?slug=bountybench · How It Works · Data refreshed daily, snapshot 2026-09-05.