Aikido CVE Rediscovery (pass@3): leaderboard
Same 26 known-CVE rediscovery set scored as a union over three runs (pass@3): a vulnerability counts as found if any of the three runs reports it, rewarding models that surface a flaw inconsistently.
Metric: Recall (%). Source: www.aikido.dev. Saturation forecast: Around January 2027. 24 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.6 Sol (Max) | 88.5 | #19 (GPT-5.6 Sol) |
| 2 | GPT-5.6 Terra (Max) | 88.5 | #41 (GPT-5.6 Terra) |
| 3 | GPT-5.6 Sol (High) | 88.5 | #19 (GPT-5.6 Sol) |
| 4 | GPT-5.6 Terra (High) | 88.5 | #41 (GPT-5.6 Terra) |
| 5 | GPT-5.5 (xHigh) | 84.6 | #31 (GPT-5.5) |
| 6 | GPT-5.6 Luna (High) | 84.6 | #68 (GPT-5.6 Luna) |
| 7 | GPT-5.6 Luna (Max) | 80.8 | #68 (GPT-5.6 Luna) |
| 8 | GPT-5.4 Mini (xHigh) | 80.8 | #212 (GPT-5.4 Mini) |
| 9 | Grok 4.5 (High) | 76.9 | #48 (Grok 4.5) |
| 10 | GPT-5.4 Mini (High) | 76.9 | #212 (GPT-5.4 Mini) |
| 11 | GPT-5.5 (High) | 73.1 | #31 (GPT-5.5) |
| 12 | Claude Opus 4.8 (xHigh) | 69.2 | #37 (Claude Opus 4.8) |
| 13 | GPT-5.4 Nano (High) | 69.2 | #328 (GPT-5.4 Nano) |
| 14 | Gemini 3.1 Pro (Preview) (High) | 65.4 | #58 (Gemini 3.1 Pro (Preview)) |
| 15 | GPT-5.4 Nano (xHigh) | 65.4 | #328 (GPT-5.4 Nano) |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=aikido-cve-rediscovery-pass-3 · How It Works · Data refreshed daily, snapshot 2026-10-20.