EuroEval Catalan Summarization - Dacsa CA: leaderboard
EuroEval's Catalan summarization task on the Dacsa CA dataset — models write abstractive summaries of Catalan news articles, scored with automatic text-overlap metrics, from the EuroEval European language model benchmark suite.
Metric: chrF3++ (%). Source: euroeval.com. Saturation forecast: Not forecast. 155 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | EuroLLM-9B-Instruct | 40.19 | #1054 |
| 2 | Mistral Small 3.1 | 39.98 | #610 |
| 3 | Apertus-70B-Instruct-2509 | 39.83 | #820 |
| 4 | Llama 3.1 8B Instruct | 39.33 | #1035 |
| 5 | occiglot-7B-eu5-instruct | 39.07 | |
| 6 | Gemma 3n E4B (IT) | 39 | #749 |
| 7 | Qwen 2.5 72B Instruct | 38.9 | #454 |
| 8 | Mistral Small 3.2 | 38.88 | #605 |
| 9 | Qwen 3 4B (Non-reasoning) | 38.84 | #829 (Qwen 3 4B) |
| 10 | Qwen 3 8B (Non-reasoning) | 38.65 | #674 (Qwen 3 8B) |
| 11 | Qwen 3 14B (Non-reasoning) | 38.64 | #537 (Qwen 3 14B) |
| 12 | Gemma 3 12B (IT) | 38.58 | #664 |
| 13 | Gemma 3 27B (IT) | 38.52 | #523 |
| 14 | Qwen 3 32B (Non-reasoning) | 38.52 | #443 (Qwen 3 32B) |
| 15 | Qwen 3 30B A3B 2507 Instruct | 38.34 | #488 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=euroeval-catalan-summarization-dacsa-ca · How It Works · Data refreshed daily, snapshot 2026-10-20.