MIR-SafetyBench - Temporal Continuity: 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 | 26.18 |
| 2 | Gemini 3 Pro (Preview) | 53.63 |
| 3 | Gemini 2.5 Pro | 61.51 |
| 4 | GPT-4o Mini | 65.62 |
| 5 | QVQ-72B-Preview | 72.24 |
| 6 | GPT-4o | 74.76 |
| 7 | Gemini 2.5 Flash | 76.34 |
| 8 | InternVL3-38B | 79.5 |
| 9 | InternVL3-8B | 79.81 |
| 10 | InternVL3-78B | 83.91 |
| 11 | Qwen 2.5 VL 32B Instruct | 85.17 |
| 12 | GLM-4.1V-9B (Thinking) | 85.49 |
Interactive version: theaggregate.ai/benchmark?slug=mir-safetybench-temporal-continuity · How It Works · Data refreshed daily, snapshot 2026-09-25.