An analysis said China has overtaken the United States for the first time in the market for open artificial intelligence (AI) models.
On the 26th, the Financial Times (FT) reported a joint release by the Massachusetts Institute of Technology (MIT) and Hugging Face, saying that last year Chinese models accounted for 17% of new global open model downloads, surpassing the 15.8% share of U.S. developers. The researchers said this is the first time China has led the United States in the field.
Open AI refers to AI technology that makes model architectures and weights freely available so anyone can download, modify, and use them. It is known to greatly help Start - Up and researchers develop products. Experts analyzed that the spread of open models is becoming a factor reshaping the global AI competitive landscape.
Since access to advanced Nvidia chips was restricted, Chinese technology corporations have, under guidance from authorities in Beijing, chosen to release models more openly. This contrasts with U.S. corporations such as OpenAI, Google, and Anthropic, which have stuck to closed strategies.
Among Chinese models, DeepSeek and Alibaba's Qwen series led in downloads. In particular, DeepSeek's reasoning model "R1" shocked Silicon Valley after earning reviews that it delivered performance similar to U.S. rivals with fewer computing resources.
MIT researchers analyzed that Chinese corporations have adopted a strategy of releasing derivative models weekly, rather than on half-year or yearly cycles like U.S. corporations. They said this approach has greatly accelerated ecosystem expansion.
Experts said that while U.S. chip export controls are limiting China's computing power, it is compensating by actively using efficiency techniques such as model distillation, backed by a large pool of researchers. They added that Chinese labs are also quickly gaining traction in AI video generation.
However, the researchers found that China's open models show a bias favorable to the Chinese Communist Party and tend to avoid answering sensitive issues such as Taiwan or the Tiananmen Square crackdown. These traits suggest that the spread of open models could also affect the information consumption environment.
By contrast, U.S. labs are focusing on developing giant frontier models, and OpenAI and Google DeepMind were said to be aiming to realize artificial general intelligence (AGI). However, there were also concerns that the number of independent research institutes active in the open-source ecosystem is smaller than in China, which could weaken competitiveness.
Janet Egan, deputy director at the Center for a New American Security (CNAS), said, "The rapid expansion of China's influence in the open model field is a signal the United States must take seriously."