Exploitbench — Epoch AI
Exploitbench — Epoch AI is run by Epoch AI. It has ranked 20 models, of which the catalogue holds 8, scoring from 0.133 to 0.474 on Mean capability.
measured by Epoch AI · the board itself
| Place | Model | Metric | Score |
|---|---|---|---|
| 3rdof 20 | GPT-5.5 | Mean capability | 0.474 |
| 7thof 20 | Claude Opus 4.7 | Mean capability | 0.265 |
| 8thof 20 | Gemini 3.1 Pro Preview | Mean capability | 0.261 |
| 10thof 20 | Claude Sonnet 4.6 | Mean capability | 0.236 |
| 12thof 20 | Kimi K2.6 | Mean capability | 0.184 |
| 13thof 20 | GLM 5.1 | Mean capability | 0.181 |
| 17thof 20 | Claude Haiku 4.5 | Mean capability | 0.137 |
| 19thof 20 | MiniMax M2.7 | Mean capability | 0.133 |
Read from the board on 2026-09-16