Penetration Testing on Cybench
55Success RateClaude 4.5 Sonnet
Evaluation Results
| Method | Links | |
|---|---|---|
| Claude 4.5 SonnetFramework=CyAgent2025.12 | 55 | |
| ARTEMISFramework=ARTEMIS2025.12 | 48.6 | |
| OpenAI GPT-5Framework=A1 (Supervisor and sub-agents)2025.12 | 45.9 | |
| Claude 4.1 OpusFramework=CyAgent2025.12 | 38 | |
| Claude 4 OpusFramework=CyAgent2025.12 | 38 | |
| Claude 4 SonnetFramework=CyAgent2025.12 | 35 | |
| OpenAI o3-miniFramework=CyAgent2025.12 | 22.5 |