CyberGym-E2E S1 (920 tasks)
Agents · 2026-06-03
CyberGym-E2E S1 success rate on the current 920-task set. A task passes S1 when the agent's generated proof of concept crashes the unpatched build.
Top models (higher is better)
| Model | Score |
|---|---|
| GPT-5.4 | 67.9 |
| Opus 4.6 | 66.3 |
| Gemini 3.1 Pro Preview | 47.4 |