Phare Misinformation (French)
Safety · 2025-02-19
Phare task testing resistance to misinformation generation or endorsement. This row is the French language split.
Top models (higher is better)
| Model | Score |
|---|---|
| Haiku 4.5 | 96.3 |
| Claude 3.7 Sonnet | 89.2 |
| Sonnet 4.5 | 87.0 |
| Opus 4.1 | 85.5 |
| Opus 4.5 | 85.5 |
| Claude 3.5 Sonnet (Oct 2024) | 83.5 |
| Llama 3.1 8B Instruct | 82.6 |
| GPT-5 Nano | 80.1 |
| GPT-5 Mini | 78.6 |
| Sonnet 5 | 77.9 |
| Gemini 3.1 Pro Preview | 77.6 |
| Opus 4.6 | 76.2 |
| Sonnet 4.6 | 74.7 |
| Qwen-Plus (2025-01-25) | 73.7 |
| Gemini 2.5 Flash-Lite | 73.5 |
| Kimi K2.6 | 73.0 |
| GPT-4.1 Nano | 72.7 |
| Haiku 3.5 | 71.5 |
| DeepSeek-V4-Pro | 71.0 |
| Gemini 1.5 Pro | 69.8 |
| DeepSeek-V4-Flash | 69.3 |
| GPT-5.1 | 68.8 |
| Gemini 3.1 Flash-Lite | 67.3 |
| Magistral Medium 1.2 | 67.3 |
| Qwen3 Max (rolling alias) | 66.1 |
| Qwen3.7 Max Preview | 65.6 |
| Grok 4.3 | 65.4 |
| Gemma 4 31B IT | 64.1 |
| GPT-4o | 63.9 |
| Gemini 3 Pro Preview | 63.4 |
| DeepSeek-V3.1 | 63.1 |
| Llama 3.1 405B Instruct | 63.1 |
| Gemini 2.5 Flash | 62.9 |
| Gemini 3.5 Flash | 62.6 |
| Mistral Large 2 (Instruct 2407) | 61.7 |
| Gemini 2.5 Pro | 61.4 |
| Llama 4 Maverick | 61.2 |
| Grok 3 Mini | 60.9 |
| Gemini 2.0 Flash | 59.7 |
| Kimi K2.5 | 59.7 |