BenCzechMark — leaderboard
Multitask Czech language benchmark evaluating LLMs across machine learning, chat, and familiarity tasks. 64+ models tested with statistical significance-based scoring.
Metric: Average Score (%). Source: huggingface.co. Status: saturated. 68 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | DeepSeek V3 (0324) | 86.4 |
| 2 | Llama 3.1 405B Instruct | 85.1 |
| 3 | DeepSeek R1 0528 | 84.82 |
| 4 | c4ai-command-a-03-2025 | 82.29 |
| 5 | Qwen 2.5 72B | 79.56 |
| 6 | MiniMax M1 80k | 76.36 |
| 7 | Llama 3.3 70B Instruct | 73.2 |
| 8 | Qwen 2.5 72B Instruct | 72.38 |
| 9 | Llama 3.1 70B Instruct | 72.3 |
| 10 | Llama 4 Scout Instruct | 71.33 |
| 11 | Llama 3.1 70B | 69.34 |
| 12 | Mistral Large 2 (Nov) Instruct (2411) | 68 |
| 13 | Qwen 2.5 32B Instruct | 67.2 |
| 14 | Gemma 4 31B (IT) | 66.12 |
| 15 | Qwen 2 72B Instruct | 64.8 |
Interactive version: theaggregate.ai/benchmark?slug=benczechmark · How the rankings work · Data refreshed daily, snapshot 2026-07-22.