Security Analysis on CyberGym
60.2Resolved PercentageSageAgent
Evaluation Results
| Method | Links | |
|---|---|---|
| SageAgentModel=GPT-5 (medium), ADK=OpenSage2026.02 | 60.2 | |
| Anthropic AgentModel=Claude Opus 4.5, ADK=Claude2026.02 | 50.6 | |
| OpenHandsModel=GPT-5 (high), ADK=OpenHands2026.02 | 39.4 | |
| Anthropic AgentModel=Claude Sonnet 4.5, ADK=Claude2026.02 | 28.9 |