LA Leaderboard - EsCoLA: leaderboard
Grammatical acceptability judgements on Spanish sentences (EsCoLA). From the LA Leaderboard, which evaluates models on Spanish, Catalan, Basque and Galician tasks with the lm-evaluation-harness; Matthews correlation x100 (signed; chance is 0), higher is better. Each model result file is a separate run.
Metric: MCC (x100). Source: huggingface.co. Saturation forecast: Estimated already saturated. 69 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemma 2 9B (IT) | 34.52 | #795 |
| 2 | Gemma 2 9B | 28.79 | #924 |
| 3 | Qwen 2.5 7B Instruct | 26.17 | #863 |
| 4 | EuroLLM-9B-Instruct | 24.57 | #1047 |
| 5 | SmolLM3-3B | 24.54 | #1239 |
| 6 | Llama 3.1 8B Instruct | 24.24 | #1017 |
| 7 | Qwen3-8B-Base | 23.06 | #626 |
| 8 | salamandra-7B | 21.46 | #1446 |
| 9 | Qwen 2.5 3B Instruct | 21.37 | #1159 |
| 10 | Qwen 2.5 7B | 20.72 | #895 |
| 11 | Gemma 3 4B (IT) | 18.96 | #997 |
| 12 | Yi-1.5-9B Chat | 18.5 | #1032 |
| 13 | EuroLLM-9B | 18.29 | #1238 |
| 14 | occiglot-7B-eu5 | 17.9 | |
| 15 | Llama 3.2 3B Instruct | 17.4 | #1327 |
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-escola · How It Works · Data refreshed daily, snapshot 2026-11-05.