RealVuln - F1: leaderboard
Metric: Strict F1 score of vulnerability findings against 796 hand-labeled entries in 26 intentionally vulnerable Python repositories, agentic scanning; higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 10 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Claude Sonnet 4.6 | 60.9 |
| 2 | Gemini 3.1 Pro (Preview) | 60.1 |
| 3 | Claude Opus 4.6 | 57.8 |
| 4 | GLM-5 | 55.5 |
| 5 | Kimi K2.5 | 54.3 |
| 6 | MiniMax-M2.7 | 48.7 |
| 7 | Claude Haiku 4.5 | 47.9 |
| 8 | Qwen 3.5 397B A17B | 46 |
| 9 | Grok 4.20 (Reasoning) | 41 |
| 10 | Grok 3 | 31.9 |
Interactive version: theaggregate.ai/benchmark?slug=realvuln-f1 · How It Works · Data refreshed daily, snapshot 2026-10-07.