MindBench - Safety Miss Rate: leaderboard
MindBench.ai safety miss rate: among chatbot replies that community raters judged clearly harmful (consensus at or below -1), the share a model rated as appropriate (+0.5 or higher). Based on about 213 eligible replies per model, so several models tie at zero; %, lower is better.
Metric: Harmful replies rated acceptable (%, lower is better). Source: mindbench.ai. Saturation forecast: Estimated already saturated. 18 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Sonnet 4.6 | 0 | #94 |
| 2 | GPT-5.4 | 0 | #83 |
| 3 | Claude Opus 4.7 | 0 | #51 |
| 4 | Gemini 3.5 Flash | 0 | #66 |
| 5 | Claude Haiku 4.5 (20251001) | 0 | #264 |
| 6 | Claude Fable 5 | 0 | #20 |
| 7 | Gemini 2.5 Pro | 0.47 | #159 |
| 8 | Gemini 2.5 Flash | 0.47 | #255 |
| 9 | GPT-5.5 | 0.47 | #33 |
| 10 | DeepSeek V4 Pro | 0.47 | #112 |
| 11 | DeepSeek V4 Flash | 0.47 | #155 |
| 12 | Sonar | 0.94 | #546 |
| 13 | O3 | 1.41 | #132 |
| 14 | Grok 4.3 | 1.41 | #204 |
| 15 | Mistral Small 4 | 1.88 | #522 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=mindbench-safety-miss-rate · How It Works · Data refreshed daily, snapshot 2026-11-05.