Antidote: Everyday Edition
Safety · 2026-05-21
Elo rating on real-world everyday prompts graded by experts.
Top models (higher is better)
| Model | Score |
|---|---|
| Claude Fable 5 | 1113 |
| Gemini 3.6 Flash | 1087 |
| Gemini 3.1 Pro Preview | 1086 |
| Gemini 3.5 Flash | 1082 |
| Kimi K3 | 1075 |
| Qwen3.7-Max | 1051 |
| Claude Opus 5 | 1049 |
| GLM-5.2 | 1045 |
| Opus 4.7 | 1041 |
| Kimi K2.6 | 1040 |
| Opus 4.6 | 1039 |
| Opus 4.8 | 1036 |
| GPT-5.6 Sol | 1021 |
| Muse Spark 1.1 | 1017 |
| Sonnet 4.6 | 1013 |
| GPT-5.5 | 1009 |
| Kimi K2.5 | 1005 |
| DeepSeek-V4-Pro | 1004 |
| Grok 4.20 | 1000 |
| Qwen3.5 Plus (2026-02-15) | 998 |
| Gemini 3.5 Flash-Lite | 997 |
| Inkling | 987 |
| Grok 4.5 | 977 |
| DeepSeek-V4-Flash | 968 |
| DeepSeek-V3.2 | 954 |
| Grok 4.3 | 953 |
| Mistral Large 3 675B Instruct 2512 | 947 |
| ERNIE 5.1 | 946 |
| Haiku 4.5 | 937 |
| GPT-5.4 Mini | 934 |
| Gemma 3 12B PT | 897 |
| ERNIE 4.5 300B A47B | 866 |
| Nova 2 Pro (Preview) | 824 |