Polish Tops List of Effective Languages for AI Models
Published on 01/11/2025 – 20:38 GMT+1
A recent study from The University of Maryland and Microsoft has shed light on which languages are most efficient for prompting artificial intelligence (AI) systems. Surprisingly, Polish emerged as the most effective language, outperforming English, which was ranked sixth among 26 languages tested.
The research involved a team of experts examining how various advanced AI models—like those from OpenAI, Google Gemini, Qwen, Llama, and DeepSeek—responded to identical prompts given in different languages. The results showed that Polish excelled, achieving an impressive accuracy rate of 88% for task completion.
The report highlighted the unexpected nature of these findings, noting that Polish, often considered a difficult language for humans to learn, proved to be the most precise when giving commands to AI. The Polish Patent Office even commented on their social media, mentioning the contrast between human difficulty with the language and the AI’s strong performance.
Interestingly, despite the limited availability of Polish training data compared to languages like English and Chinese, the AI systems demonstrated a robust understanding of Polish. On the other hand, Chinese ranked fourth from the bottom in this study, indicating that it struggled more in comparison.
Here are the top 10 languages for effective prompts to conversational AI:
- Polish: 88%
- French: 87%
- Italian: 86%
- Spanish: 85%
- Russian: 84%
- English: 83.9%
- Ukrainian: 83.5%
- Portuguese: 82%
- German: 81%
- Dutch: 80%
This study highlights the evolving relationship between language learning and artificial intelligence, revealing that AI might be more adept in some languages than even the native speakers themselves.
