That was when many of us first realised that the race for the AI crown was being fought between the United States and China. Today, however, it is worth noticing something even more important. Every such clash between technology giants drives prices down, improves quality and broadens access to advanced technologies. What does the rivalry between Fable and Kimi mean for businesses, schools and each of us? The issue is discussed by Dr Dominik Skowroński from the Faculty of Management, University of Lodz.
Déjà Vu from Across the Pacific
History has a tendency to repeat itself. In June 2026, Anthropic released Claude Fable 5, currently regarded as the world’s most capable artificial intelligence model. Five weeks later, China responded with the launch of Kimi K3. Comparisons conducted by independent analysts indicate that the American model still holds the crown, winning 22 out of 35 shared benchmark tests. At the same time, Kimi excels in tasks requiring long, independent work on software code and is approximately 70% cheaper across all pricing tiers. A year earlier, DeepSeek had played exactly the same card. At that time, markets reacted with panic. Today, they react with little more than a shrug, as this exchange of blows has become the natural rhythm of the industry.
From a user’s perspective, such competition resembles a price war between two supermarket chains. Nobody knows who will ultimately win, but everyone’s shopping basket becomes cheaper. Following the launch of DeepSeek, access costs for AI models began falling across the market, and we are now observing precisely the same mechanism once again.
Open Weights, Open Doors
The most interesting aspect of the Chinese strategy is its openness. Moonshot AI has made its model weights available for download, following the example previously set by DeepSeek. In practice, this means that a university, hospital or public institution can deploy a powerful AI model on its own servers without sending sensitive data overseas. It is rather like the difference between renting a house and being given the architectural plans to build and customise one yourself. For institutions that take digital sovereignty seriously, open models are becoming a genuine alternative. Poland also benefits, as developers of domestic models, such as Bielik and PLLuM, can draw upon solutions published by the largest international players.
Poland: Greater Potential than Actual Adoption
Against this backdrop, the Polish reality appears rather modest. According to Statistics Poland (GUS), 8.7% of companies in Poland used AI technologies in 2025, compared with 5.9% the previous year. The increase is significant, yet more than 90% of enterprises remain on the sidelines of a revolution unfolding before their eyes. The barrier is no longer financial. As access costs to the most advanced models continue to fall, and some models are available free of charge, the real bottleneck has become AI competence. During the training sessions I conduct, I repeatedly observe the same pattern. People do not need lectures about model parameters. They need practical training that demonstrates how to translate the capabilities of artificial intelligence into solutions for their own tasks and challenges.
Who Really Wins This Race?
The question of who will ultimately sit on the AI throne will be decided in laboratories in California and Beijing. The question of who benefits from the results of this race is decided by people and organisations capable of turning access to technology into the ability to use it effectively. The greatest advantage will belong to companies and educational institutions that transform the way they work, not those that simply purchase the largest number of tools. Cheering for the giants will bring us little benefit. Instead, we should take three practical steps: test the latest models on real-world tasks; plan the development of AI-related competencies within teams; identify areas where AI can relieve people of repetitive and time-consuming work. The giants are fighting for the throne. The rewards are being distributed among the audience.
Author: Dr Dominik Skowroński
Edit: Faculty of Management, University of Lodz
