MIR-SafetyBench - Temporal Jump: leaderboard
Metric: Attack Success Rate (%; lower is safer). Source: arxiv.org. Saturation forecast: Around December 2026. 19 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.1 | 17.49 |
| 2 | Gemini 3 Pro (Preview) | 44.88 |
| 3 | GPT-4o Mini | 55.78 |
| 4 | Gemini 2.5 Pro | 58.42 |
| 5 | GPT-4o | 67.66 |
| 6 | InternVL3-78B | 71.62 |
| 7 | Gemini 2.5 Flash | 73.27 |
| 8 | QVQ-72B-Preview | 75.91 |
| 9 | InternVL3-8B | 77.23 |
| 10 | InternVL3-38B | 82.84 |
| 11 | GLM-4.1V-9B (Thinking) | 86.47 |
| 12 | Qwen 2.5 VL 32B Instruct | 88.12 |
Interactive version: theaggregate.ai/benchmark?slug=mir-safetybench-temporal-jump · How It Works · Data refreshed daily, snapshot 2026-09-25.