Compare AI models
| Benchmarks | ||
CyberGymCybersecurity — Tests the AI on cybersecurity challenges — finding and exploiting software weaknesses inside a safe sandbox. Higher is better. | 86.2% | 85.6% |
| Overview | ||
| Company | OpenAI | |
| Release date | Sep 2 2026 | Jun 22 2026 |
| Access | Closed | Closed |
| Model details | View model | View model |
Other comparisons
Gemini 3.8 Flash CybervsClaude Haiku 5.5GPT-5.5-CybervsClaude Haiku 5.5Gemini 3.8 Flash CybervsMuse Spark OpenGPT-5.5-CybervsMuse Spark OpenGemini 3.8 Flash CybervsGrok 4.7GPT-5.5-CybervsGrok 4.7Gemini 3.8 Flash CybervsDeepSeek-V4.1-Flash-1017GPT-5.5-CybervsDeepSeek-V4.1-Flash-1017Gemini 3.8 Flash CybervsMistral Large 4GPT-5.5-CybervsMistral Large 4Gemini 3.8 Flash CybervsKimi K3GPT-5.5-CybervsKimi K3Frequently asked questions
Gemini 3.8 Flash Cyber leads GPT-5.5-Cyber on 1 of the 1 benchmark they both report (CyberGym). GPT-5.5-Cyber shipped 72 days before Gemini 3.8 Flash Cyber, so benchmark comparisons should account for the intervening progress.
Published specifications for these two models are limited — see each model page for the latest details.
On CyberGym, Gemini 3.8 Flash Cyber leads at 86.2% vs GPT-5.5-Cyber at 85.6%.