EuroEval French Summarization - Orange SUM: leaderboard
EuroEval's French summarization task on the Orange SUM dataset — models write abstractive summaries of French news articles, scored with automatic text-overlap metrics, from the EuroEval European language model benchmark suite.
Metric: chrF3++ (%). Source: euroeval.com. Saturation forecast: Around 2034. 151 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Bielik-11B-v2.3-Instruct | 39.96 | #693 |
| 2 | Gemini 3 Pro (Preview) | 39.53 | #66 |
| 3 | Claude Sonnet 4.5 (Non-reasoning) | 39.5 | #142 (Claude Sonnet 4.5) |
| 4 | Qwen 3.5 35B A3B Base | 39.12 | #310 |
| 5 | Gemma 4 26B A4B (IT) | 39.1 | #297 |
| 6 | Gemma 4 31B (IT) | 38.97 | #176 |
| 7 | GPT-4.1 | 38.87 | #248 |
| 8 | Gemini 3.1 Flash Lite (Preview) | 38.63 | #194 |
| 9 | Gemini 2.5 Flash Lite (Non-reasoning) | 38.62 | #422 (Gemini 2.5 Flash Lite) |
| 10 | Gemini 3 Flash (Preview) (Non-reasoning) | 38.62 | #82 (Gemini 3 Flash (Preview)) |
| 11 | Mistral Small 3.1 | 38.55 | #610 |
| 12 | occiglot-7B-eu5-instruct | 38.47 | |
| 13 | EuroLLM-9B-Instruct | 38.45 | #1054 |
| 14 | Gemini 3 Flash (Preview) (Thinking) | 38.41 | #82 (Gemini 3 Flash (Preview)) |
| 15 | Llama 3.3 70B Instruct | 38.39 | #534 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=euroeval-french-summarization-orange-sum · How It Works · Data refreshed daily, snapshot 2026-10-20.