Agents · Benchmark profile
A cybersecurity-agent benchmark measuring progress from reaching vulnerable code through exploit primitives and code execution.
Data verified 15 Jul 2026 · Methodology 1.8.0
Benchmark score on ExploitBench
GPT-5.5 leads at 34%, followed by Claude Opus 4.7 (27%).
2 ranked models · higher is better · labels show rank and score
| Model | Rank | Provider | Score |
|---|---|---|---|
| GPT-5.5 | #1 | OpenAI | 34% |
| Claude Opus 4.7 | #2 | Anthropic | 27% |
One best score per model · higher is better
| Rank | Model | Provider | License | Evidence use | Score |
|---|---|---|---|---|---|
| #1 | GPT-5.5 gpt-5.5 | OpenAI | closed | Ranking eligible | 34% |
| #2 | Claude Opus 4.7 claude-opus-4-7 | Anthropic | closed | Ranking eligible | 27% |
About ExploitBench
A cybersecurity-agent benchmark measuring progress from reaching vulnerable code through exploit primitives and code execution. Results stay tied to the exact model variant and evaluation system. Multiple systems for the same model use the best published score on this page; overall Lumina scoring uses the median of ranking-eligible rows.
Open benchmark source ↗FAQ
A cybersecurity-agent benchmark measuring progress from reaching vulnerable code through exploit primitives and code execution.
GPT-5.5 by OpenAI currently leads with 34%.
2 models in the LuminaBench cohort have a qualifying score on this benchmark.
Yes. This family is ranking-weighted at 1.6% of its capability category. GPT-5.5 is overall #14.
Related
