Safety Evaluation on Production Benchmarks
92.6Hate (Non-Violent)gpt-5-thinking-nano
Evaluation Results
| Method | Links | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| gpt-5-thinking-nano2025.12 | 92.6 | 90.9 | 77.1 | 93.5 | 97.2 | 95.5 | 79.7 | 87.9 | 94.7 | 93.3 | 95.4 | |
| gpt-5-thinking-mini2025.12 | 87.4 | 84.3 | 75.2 | 94 | 98.1 | 94.4 | 82.9 | 81.4 | 94.4 | 95 | 93.9 | |
| OpenAI o4-mini2025.12 | 87.4 | 84.7 | 64.1 | 92.7 | 94.7 | 86.4 | 72.4 | 60.3 | 78.6 | 81.5 | 86.4 | |
| gpt-5-main-mini2025.12 | 82.1 | 94.3 | 70.8 | 88.6 | 92.9 | 90.9 | 77.8 | 75.6 | 80.9 | 86.4 | 77.3 |