Phare Factuality (French)
Safety · 2025-02-19
Phare task testing factual answer reliability and hallucination resistance. This row is the French language split.
Top models (higher is better)
| Model | Score |
|---|---|
| Gemini 3.1 Pro Preview | 82.3 |
| GPT-5.5 | 78.9 |
| Gemini 3.5 Flash | 78.2 |
| Gemini 3 Pro Preview | 76.5 |
| Claude 3.5 Sonnet (Oct 2024) | 73.8 |
| GPT-5 | 73.5 |
| Grok 3 | 73.5 |
| Grok 4 | 73.5 |
| GPT-5.2 | 72.1 |
| Opus 4.5 | 71.9 |
| Opus 4.6 | 71.8 |
| Gemini 2.5 Pro | 71.8 |
| GPT-4.1 | 71.1 |
| Claude 3.7 Sonnet | 70.8 |
| GPT-4o | 70.8 |
| GPT-5.1 | 70.8 |
| Kimi K2.5 | 70.8 |
| Gemini 3.1 Flash-Lite | 69.7 |
| Kimi K2.6 | 69.7 |
| Sonnet 5 | 68.4 |
| DeepSeek-V4-Pro | 68.4 |
| Grok 4.3 | 68.4 |
| Sonnet 4.6 | 68.3 |
| DeepSeek-V3-0324 | 68.0 |
| DeepSeek-V4-Flash | 68.0 |
| Opus 4.1 | 67.3 |
| Gemini 1.5 Pro | 67.2 |
| Gemini 2.0 Flash | 67.0 |
| Qwen3.7 Max Preview | 67.0 |
| GLM-5.2 | 66.3 |
| DeepSeek-V3 | 66.0 |
| Sonnet 4.5 | 65.5 |
| Mistral Large 2 (Instruct 2407) | 64.3 |
| Qwen3.7-Plus | 62.9 |
| Mistral Large 3 675B Instruct 2512 | 62.6 |
| DeepSeek-R1-0528 | 62.2 |
| GPT-5 Mini | 62.2 |
| Mistral Medium 3.1 | 62.1 |
| Grok 3 Mini | 61.9 |
| Gemini 2.5 Flash | 61.6 |