French LLM Leaderboard - Average: leaderboard
French leaderboard headline averages the normalized IFEval FR, GPQA FR and BAC FR scores used by the official frontend. GPQA normalization removes the chance baseline; separate task boards preserve raw measurements.
Metric: Mean Normalized Score (%). Source: huggingface.co. Saturation forecast: Estimated already saturated. 55 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | DeepSeek R1 Distill Llama 70B | 55.93 | #594 |
| 2 | Mistral Large 2 (Nov) Instruct (2411) | 49.41 | #465 |
| 3 | Llama 3.3 70B Instruct | 48.54 | #534 |
| 4 | DeepSeek R1 Distill Qwen 32B | 48.32 | #655 |
| 5 | Qwen 2.5 72B Instruct | 46.93 | #454 |
| 6 | calme-3.2-instruct-78B | 43.34 | #347 |
| 7 | Phi-4 | 42.27 | #706 |
| 8 | Qwen 2.5 14B Instruct | 41.86 | #647 |
| 9 | Falcon3-10B-Instruct | 38.28 | #1157 |
| 10 | Mistral Small 3 | 37.98 | #627 |
| 11 | Mixtral 8x22B Instruct (v0.1) | 37.67 | #756 |
| 12 | Falcon3-7B-Instruct | 36.01 | #1305 |
| 13 | Phi-3.5-MoE-instruct | 32.67 | #700 |
| 14 | Phi-3-medium-128k-instruct | 29.03 | #1013 |
| 15 | granite-3.1-8B-instruct | 27.25 | #1069 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=french-llm-leaderboard-average · How It Works · Data refreshed daily, snapshot 2026-10-20.