# CyberGym — AI model rankings

Tests the AI on cybersecurity challenges — finding and exploiting software weaknesses inside a safe sandbox. Higher is better.

13 tracked models have a published CyberGym score. Higher is better. Scores come from published lab reports and benchmark sources; recorded sources appear beside the scores.

## Ranking

| Rank | Model | Developer | Score | Source | Released |
| --- | --- | --- | --- | --- | --- |
| 1 | DeepSeek-V4.1-Flash | DeepSeek | 88.1% | Lab | Sep 10 2026 |
| 2 | Gemini 3.8 Flash Cyber | Google | 86.2% | Lab | Sep 2 2026 |
| 3 | GPT-5.5-Cyber | OpenAI | 85.6% | Lab | Jun 22 2026 |
| 4 | GLM-5.3 | Z.ai | 84.5% | Lab | Aug 14 2026 |
| 5 | DeepSeek-V4-Pro-0813 | DeepSeek | 83.3% | Lab | Aug 13 2026 |
| 6 | GPT-5.5 | OpenAI | 81.8% | Lab | Apr 23 2026 |
| 7 | Grok 4.7 | SpaceXAI | 80.3% | Lab | Sep 21 2026 |
| 8 | Grok 4.6 | SpaceXAI | 79.7% | Lab | Aug 12 2026 |
| 9 | GPT-5.4 | OpenAI | 79% | Lab | Mar 5 2026 |
| 9 | Grok 4.5 | SpaceXAI | 79% | Lab | Jul 8 2026 |
| 11 | Gemini 3.5 Flash Cyber | Google | 77.5% | Lab | Jul 21 2026 |
| 12 | DeepSeek-V4-Flash-0731 | DeepSeek | 76.7% | Lab | Jul 31 2026 |
| 13 | Claude Opus 4.7 | Anthropic | 73.1% | Lab | Apr 16 2026 |


---

Canonical page: https://aireleasetracker.com/benchmark/cybergym
Full dataset: https://aireleasetracker.com/llms-full.txt · JSON: https://aireleasetracker.com/models.json
Source: AI Release Tracker (https://aireleasetracker.com). Most benchmark scores come from lab launch material; gathered results identify the leaderboard that published them.