ExploitBench v8-bench: leaderboard
ExploitBench evaluates agents on patched V8 vulnerabilities. Capability coverage (%), all-runs view with reported seed/turn budgets; baseline. This is capability-flag coverage, not mean capability count. Baseline and AutoNudge prompting are separate boards; each source run retains its scaffold and reported seed/turn budget. Scores use the published all-runs view.
Metric: Capability coverage (%), all-runs view with reported seed/turn budgets; baseline. Source: exploitbench.ai. Saturation forecast: Around January 2027. 10 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Mythos Preview | 72 | #4 |
| 2 | GPT-5.5 | 42 | #31 |
| 3 | Gemini 3.1 Pro (Preview) | 26 | #58 |
| 4 | Claude Sonnet 4.6 | 25 | #90 |
| 5 | Claude Opus 4.7 | 25 | #49 |
| 6 | Kimi K2.6 | 16 | #104 |
| 7 | GLM-5.1 | 16 | #97 |
| 8 | Claude Haiku 4.5 | 14 | #280 |
| 9 | MiniMax-M2.7 | 13 | #258 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=exploitbench-v8-bench · How It Works · Data refreshed daily, snapshot 2026-10-20.