Open AI models from China put European businesses before a choice

Open source artificial intelligence models are rapidly gaining popularity among businesses and developers seeking cost-effective alternatives that reduce reliance on major American tech labs. This shift has accelerated due to skyrocketing token costs, recent American export bans on models like Anthropic's Fable 5, and simultaneous service outages among top US providers.

Chinese AI companies have capitalized on this trend by releasing highly capable open models such as Z.ai 5.3, Moonshot's Kimi K3, Alibaba's Qwen 3.8, and Deepseek. These models now surpass proprietary alternatives in usage on the Open Router marketplace. Organizations choose them for lower prices, technological sovereignty, and the ability to run them on their own hardware or preferred cloud services, avoiding dependency on both American and European infrastructure.

While American companies offer competing open models from firms like Meta and Nvidia, Europe has largely stepped back from developing frontier models, with leaders instead focusing on practical AI applications. Meanwhile, local concerns in Sweden, such as fears surrounding data handling and controversies in the public sector, often push the use of Chinese AI into secrecy.

Pragmatically, running these models locally or through trusted cloud services prevents data from being sent to China, and their open weights allow for independent security analysis. Ultimately, decision-makers face a geopolitical choice between adopting affordable Chinese technology that challenges American dominance or relying on costly and potentially volatile American platforms.