MIR-SafetyBench - Logical Analogy: leaderboard
Metric: Attack Success Rate (%; lower is safer). Source: arxiv.org. Saturation forecast: Estimated already saturated. 19 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.1 | 5.03 |
| 2 | Gemini 3 Pro (Preview) | 26.1 |
| 3 | Gemini 2.5 Pro | 27.36 |
| 4 | Gemini 2.5 Flash | 42.77 |
| 5 | GPT-4o Mini | 55.35 |
| 6 | GPT-4o | 58.49 |
| 7 | InternVL3-78B | 67.3 |
| 8 | QVQ-72B-Preview | 67.92 |
| 9 | InternVL3-8B | 78.62 |
| 10 | InternVL3-38B | 81.13 |
| 11 | Qwen 2.5 VL 32B Instruct | 81.76 |
| 12 | GLM-4.1V-9B (Thinking) | 93.4 |
Interactive version: theaggregate.ai/benchmark?slug=mir-safetybench-logical-analogy · How It Works · Data refreshed daily, snapshot 2026-09-25.