The first 'Polish' test for large language models (LLMs) has been introduced, with results that were surprisingly unexpected. The test comprised 20 tasks across 10 categories, focusing on practical skills such as email writing, advice for entrepreneurs, legal and tax support, as well as aspects of Polish culture including quoting the first twelve lines of 'Pan Tadeusz' and identifying Christmas customs. Given this diverse set, one might assume that Polish artificial intelligence would perform better, but the findings suggest otherwise.

Publishing the results for the first time, 'Rzeczpospolita' highlighted that Google's tool dominated the rankings, demonstrating its superior capabilities. Surprisingly, the Chinese model Qwen secured a place on the podium, outpacing the LLM developed by Meta. This indicates a significant gap in performance between Polish-developed models and those from leading tech giants, underscoring the need for improvement in the Polish AI sector.

These results not only point to the competitive edge of advanced AI development in countries like the USA and China but also challenge Polish developers to enhance their technology. The implications for the future of AI in Poland and its potential applications across various sectors may depend on a concerted effort to innovate and close the gap with the globally leading models.