Aikido Known-CVE Detection (32 CVEs) single-run average
Code · 2026-10-22
Mean single-run recall across three independent audits of Aikido's expanded 32-CVE known-vulnerability set (October 2026); a one-run consistency diagnostic beside the pass@3 headline.
Top models (higher is better)
| Model | Score |
|---|---|
| DeepSeek V4.1 Flash 1017 | 70.8 |
| Claude Opus 5 | 70.8 |
| Grok 4.7 | 70.8 |
| DeepSeek V4.1 Flash | 65.6 |