Multimodal Safety Evaluation on ChatGPT image input safety evaluations
98.9Hate SafetyGPT-4o
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| GPT-4o2025.12 | 98.9 | 96.4 | 94.6 | 95.6 | 98 | 99.5 | |
| gpt-5-main2025.12 | 98.6 | 99.1 | 98.6 | 100 | 99.7 | 99.4 | |
| gpt-5-thinking2025.12 | 96.8 | 98 | 98.8 | 100 | 99.6 | 99.4 | |
| OpenAI o32025.12 | 93.5 | 96.2 | 97.2 | 98 | 98.2 | 98.7 |