Exploitbench
The Frontier
Best score over time · one chart, every benchmark
Full rankings
8 models tested · sorted by score
| # | Model | Score |
|---|---|---|
| 1 | 47.4 | |
| 2 | 26.5 | |
| 3 | 26.1 | |
| 4 | 23.6 | |
| 5 | 18.4 | |
| 6 | 18.1 | |
| 7 | 13.7 | |
| 8 | 13.3 |
What it measures · how it is scored · sources · as of 2026-11-17
- How it is scored · Scored as a percentage · higher is better · maximum 100
- Released · 2026-05-13
- Scores from · Epoch AI Benchmarking Hub (8 scores)
Score distribution
Where models cluster
Correlated benchmarks
Pearson r · original research
Benchmarks that track with Exploitbench
Pearson correlation across models scored on both benchmarks. Closer to 1 = strongly predictive.
Frequently asked
About Exploitbench
What does Exploitbench measure?
Exploitbench is a knowledge benchmark in the BenchGecko catalog. 8 AI models have been tested on it. Scores range from 13.3 to 47.4 out of 100.
Which model leads on Exploitbench?
GPT-5.5 from OpenAI leads Exploitbench with a score of 47.4. The median score across 8 tested models is 21.0.
Is Exploitbench saturated?
No · the top score is 47.4 out of 100 (47%). There is still meaningful room for improvement on Exploitbench.
Does Exploitbench predict performance on other benchmarks?
Yes · Exploitbench scores correlate 0.95 with Mystery Game Puzzles across 5 shared models. Models that do well on Exploitbench tend to do well on Mystery Game Puzzles.
How often is Exploitbench data refreshed?
BenchGecko pulls updates daily. New model scores on Exploitbench appear as soon as they are published by Epoch AI or the model provider.
- Category
- Knowledge
- Max score
- 100
- Models
- 8
- Updated
- 2026-04-24
Top on Exploitbench
GPT-5.5 · 47.4Claude Opus 4.7 · 26.5Gemini 3.1 Pro Preview · 26.1Claude Sonnet 4.6 · 23.6Kimi K2.6 · 18.4More knowledge benchmarks
Same category · related evaluations