EuroEval English NLU - SQuAD: leaderboard
The "SQuAD" subtask of EuroEval English NLU. English-language EuroEval natural-language-understanding suite, separating NLU task performance from the broader all-task EuroEval aggregate.
Metric: Character F1 (%). Source: euroeval.com. Saturation forecast: Estimated already saturated. 327 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Llama 3.1 405B | 89.95 | #466 |
| 2 | Qwen 3 32B (Non-reasoning) | 89.16 | #443 (Qwen 3 32B) |
| 3 | Ministral-3-14B-Reasoning-2512 | 88.68 | #540 |
| 4 | Magistral Small 1.2 | 88.59 | #570 |
| 5 | Yi 1.5 34B | 88.19 | #807 |
| 6 | Gemma 2 27B | 88.11 | #806 |
| 7 | Mistral Small 3.2 | 87.91 | #605 |
| 8 | OLMo-2-1124-7B | 87.82 | #1401 |
| 9 | Llama 3.1 70B | 87.68 | #592 |
| 10 | aya-23-35B | 87.4 | #883 |
| 11 | Qwen 3 14B (Non-reasoning) | 87.26 | #537 (Qwen 3 14B) |
| 12 | Voxtral-Small-24B-2507 | 87.05 | #581 |
| 13 | Llama 2 70B Base | 86.97 | #1025 |
| 14 | Yi-1.5-9B | 86.89 | #1241 |
| 15 | Qwen3.5-2B-Base | 86.3 | #1309 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=euroeval-english-nlu-squad · How It Works · Data refreshed daily, snapshot 2026-10-20.