Cybersecurity
CyberGym
Tests the AI on cybersecurity challenges — finding and exploiting software weaknesses inside a safe sandbox. Higher is better.
Rankings
Higher is betterCyberGym — frequently asked questions
- What is CyberGym?
- Tests the AI on cybersecurity challenges — finding and exploiting software weaknesses inside a safe sandbox. Higher is better.
- Which AI model scores highest on CyberGym?
- GPT-5.5 by OpenAI holds the best CyberGym result among tracked models, at 81.8% (released Apr 23 2026). Higher scores are better on this benchmark.
- What are the top 4 models on CyberGym?
- 1. GPT-5.5 (OpenAI) — 81.8%; 2. GPT-5.4 (OpenAI) — 79%; 3. DeepSeek-V4-Flash-0731 (DeepSeek) — 76.7%; 4. Claude Opus 4.7 (Anthropic) — 73.1%.
- How many models have a published CyberGym score?
- 4 tracked models have a published CyberGym score. Scores are the figures reported by each lab at that model's release, so this page is a record of results over time rather than a re-run leaderboard.