The results of research conducted by American scientists from the University of Maryland and Microsoft have revealed a surprising “advantage” of the Polish language. In the latest report, it was placed first in terms of performance of rich, contextual prompts for artificial intelligence.
See also: Record money in speculative funds betting on AI. There is a change
The results of the analysis show that the Polish language, despite relatively small data resources for training AI models, may be more effective in the context of long and complex commands than languages with a larger number of users. This discovery may be important for the development of AI technology in Poland and around the world, pointing to the potential of smaller languages in a global context.
The study cited by Benchmark.pl, titled “One ruler to measure them all: Benchmarking multilingual long-context language models”, covered 26 languages and several AI models, including OpenAI o3-mini-high, Google Gemini 1.5 Flash and Llama 3.3 (70B). The researchers tested the models' performance in processing prompts up to 128,000 words long. tokens. The results were surprising – Polish achieved the highest effectiveness (88%), ahead of, among others, French (87%) and Italian (86%).
Polish is the best language for AI. English only ranks sixth
“Experiments with open and closed LLM models reveal an increasing difference in performance between low- and high-resource languages as the context length increases from 8K to 128K tokens. Surprisingly, English is not the language with the highest performance on long context tasks (it took 6th place out of 26), and the Polish language comes to the top of the list,” it was written in the report, quoted by Benchmark.pl.
See also: This is not what social media was supposed to look like. AI is pushing people out of the loop
The full list of language performance in the context of long prompts is as follows:
- Poland – 88 percent
- French – 87 percent
- Italian – 86 percent
- Spanish – 85 percent
- Russian – 84 percent
- English – 83.9 percent
The study results may be an impulse for greater involvement in the development of AI technology in Poland. They show that the Polish language, despite its complexity, can be an asset in the global technological race. It's also proof that smaller languages can play a key role in the future of artificial intelligence.
