LA Leaderboard - RagQuAS: leaderboard
Spanish retrieval-augmented question answering over supplied passages (RagQuAS). From the LA Leaderboard, which evaluates models on Spanish, Catalan, Basque and Galician tasks with the lm-evaluation-harness; semantic answer similarity, the cosine similarity of sentence embeddings x100, higher is better. Each model result file is a separate run.
Metric: Semantic Answer Similarity (x100). Source: huggingface.co. Saturation forecast: Estimated already saturated. 69 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | EuroLLM-9B-Instruct | 81.58 | #1047 |
| 2 | Llama 3.1 8B | 81.39 | #1130 |
| 3 | Qwen 2.5 1.5B | 81.33 | #1412 |
| 4 | Llama 3.2 3B | 81.1 | #1376 |
| 5 | Qwen 2.5 3B | 80.86 | #1177 |
| 6 | Qwen 2.5 7B | 80.85 | #895 |
| 7 | Teuken-7B-Instruct-v0.4 | 80.63 | #1408 |
| 8 | SmolLM3-3B-Base | 80.6 | #1260 |
| 9 | Teuken-7B-instruct-commercial-v0.4 | 80.5 | #1421 |
| 10 | Mistral 7B Instruct (v0.3) | 80.39 | #1255 |
| 11 | Qwen3-8B-Base | 80.35 | #626 |
| 12 | SmolLM2-1.7B-Instruct | 80.05 | #1541 |
| 13 | occiglot-7B-eu5-instruct | 79.64 | |
| 14 | Yi-1.5-9B Chat | 79.51 | #1032 |
| 15 | Llama 3.2 1B | 78.96 | #1540 |
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-ragquas · How It Works · Data refreshed daily, snapshot 2026-11-05.