Benchmarks
Exploitbench
Models scored
19
Best score
0.69
Metric
Mean capability
Exploitbench is an AI evaluation tracked by Epoch AI, with 19 scored model versions on record. The scores here are reported by third parties rather than produced by Epoch AI running the evaluation itself.
Scored models span October 2025 to April 2026. The best score recorded is 0.69, measured as Mean capability. Models come from Anthropic, OpenAI, Google DeepMind, Moonshot, Z.ai (Zhipu AI), MiniMax.
Full record
- Score metric
- Mean capability
- Models scored
- 19
- Best score recorded
- 0.69
- Earliest model scored
- Oct 15, 2025
- Most recent model scored
- Apr 23, 2026
- Third-party reported
- Yes
Leaderboard
| 01 | claude-mythos-preview-early | Anthropic | 0.69 |
| 02 | gpt-5.5_unknown | OpenAI | 0.41 |
| 03 | claude-opus-4-7 | Anthropic | 0.28 |
| 04 | claude-opus-4-7_unknown | Anthropic | 0.27 |
| 05 | gemini-3.1-pro-preview | Google DeepMind | 0.26 |
| 06 | claude-sonnet-4-6_unknown | Anthropic | 0.24 |
| 07 | glm-5.1 | Z.ai (Zhipu AI) | 0.18 |
| 08 | kimi-k2.6 | Moonshot | 0.18 |
| 09 | claude-haiku-4-5-20251001_unknown | Anthropic | 0.14 |
| 10 | MiniMax-M2.7 | MiniMax | 0.13 |
More benchmarks
SourceEpoch AI, 'AI Benchmarking Hub'. Published online at epoch.ai. Retrieved 2026-07-29 from https://epoch.ai/benchmarks. Licensed under CC BY 4.0.
← All benchmarks