EuroEval Czech NLU - SQAD: leaderboard
The "SQAD" subtask of EuroEval Czech NLU. Czech-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: Around January 2027. 228 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Ministral-3-14B-Reasoning-2512 | 72.74 | #540 |
| 2 | Qwen3.5-4B-Base | 72.57 | #798 |
| 3 | GPT-5 | 72.33 | #98 |
| 4 | GPT-5.4 Mini (2026-03-17) (High) | 70.77 | #169 (GPT-5.4 Mini (2026-03-17)) |
| 5 | GPT-5 Mini (2025-08-07) | 70.67 | #172 |
| 6 | Qwen 3.5 35B A3B FP8 | 70.3 | #362 |
| 7 | GPT-5 Nano | 70.09 | #425 |
| 8 | SOLAR-10.7B-v1.0 | 69.78 | #1253 |
| 9 | Ministral-3-14B-Instruct-2512 | 69.7 | #615 |
| 10 | gemma-3-12B-pt | 69.39 | #686 |
| 11 | Claude Sonnet 4.5 (Non-reasoning) | 69.2 | #142 (Claude Sonnet 4.5) |
| 12 | gemma-3-27B-pt | 68.88 | #502 |
| 13 | Llama 3.1 70B | 68.83 | #592 |
| 14 | GPT-5.2 | 68.56 | #110 |
| 15 | Magistral Small 1.2 | 68.49 | #570 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=euroeval-czech-nlu-sqad · How It Works · Data refreshed daily, snapshot 2026-10-20.