LA Leaderboard - XL-Sum Spanish: leaderboard
Summarisation of Spanish BBC news articles from XL-Sum. From the LA Leaderboard, which evaluates models on Spanish, Catalan, Basque and Galician tasks with the lm-evaluation-harness; sacreBLEU (0-100), higher is better. Each model result file is a separate run.
Metric: BLEU. Source: huggingface.co. Saturation forecast: Not forecast. 68 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | aya-expanse-8B | 4.33 | #914 |
| 2 | Gemma 2 9B (IT) | 3.98 | #795 |
| 3 | EuroLLM-9B | 3.77 | #1238 |
| 4 | OLMo-2-1124-7B | 3.59 | #1381 |
| 5 | Yi-1.5-9B | 3.26 | #1215 |
| 6 | occiglot-7B-eu5 | 2.88 | |
| 7 | Mistral-7B-v0.3 | 2.88 | #1321 |
| 8 | Gemma 2 9B | 2.85 | #924 |
| 9 | Qwen 2.5 3B Instruct | 2.65 | #1159 |
| 10 | Llama 3.1 8B Instruct | 2.62 | #1017 |
| 11 | Gemma 2 2B (IT) | 2.58 | #1328 |
| 12 | Teuken-7B-instruct-commercial-v0.4 | 2.46 | #1421 |
| 13 | Qwen 2.5 7B Instruct | 2.39 | #863 |
| 14 | EuroLLM-9B-Instruct | 2.34 | #1047 |
| 15 | SmolLM3-3B | 2.33 | #1239 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=la-leaderboard-xl-sum-spanish · How It Works · Data refreshed daily, snapshot 2026-11-05.