ExploitBench June-August 2026 - ACE
Code
20 recent high-severity V8 vulnerabilities across 13 stable Chrome releases. Measures arbitrary code execution on V8 and official Linux Chrome under a 300-turn evaluation limit; release scores average repeated trials.
Top models (higher is better)
| Model | Score |
|---|---|
| GPT-6 Astra | 39.0 |