Gemini 3.8 Flash Cyber: benchmark results
Google's security-tuned Gemini 3.8 Flash Cyber (September 2, 2026), available only to vetted defenders through the limited-access Fairwind Program. Provider: Google. Released 2026-09-02. Access: Private.
Aggregate Rating 1885 ± 32, rank #26 of 10916 rated models, from 19 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Wiz Cyber Model Arena (ADK React) - Websec.fr CTFs | 90.8 | Websec.fr CTFs challenges solved, pass@1 (%) | 100 |
| Wiz Cyber Model Arena (Antigravity) - Websec.fr CTFs | 77 | Websec.fr CTFs challenges solved, pass@1 (%) | 100 |
| Wiz Cyber Model Arena (Claude Code) - API Security | 84.6 | API Security challenges solved, pass@1 (%) | 100 |
| Wiz Cyber Model Arena (ADK React) - API Security | 85.2 | API Security challenges solved, pass@1 (%) | 95 |
| Wiz Cyber Model Arena (ADK React) - Overall | 74.9 | Macro average of the four category pass@1 rates (%) | 95 |
| CWE-Bench | 47.2 | Pass@1 (%) | 92.9 |
| Wiz Cyber Model Arena (ADK React) - Cloud Security | 66.7 | Cloud Security challenges solved, pass@1 (%) | 92.5 |
| Wiz Cyber Model Arena (ADK React) - Code Vulnerabilities | 57 | Code Vulnerabilities challenges solved, pass@1 (%) | 90 |
| LLM Stats (CyberGym) | 86.2 | Score (%) | 85 |
| LLM Stats Score | 39.11 | LLM Stats Score (conservative rating) | 81 |
| Wiz Cyber Model Arena (Claude Code) - Overall | 68.5 | Macro average of the four category pass@1 rates (%) | 79.2 |
| Wiz Cyber Model Arena (Antigravity) - API Security | 84 | API Security challenges solved, pass@1 (%) | 75 |
Interactive version: theaggregate.ai/model?slug=gemini-3-8-flash-cyber · How It Works · Data refreshed daily, snapshot 2026-11-08.