China's Open-Weight AI Models Overtake US in Global Downloads for First Time — MIT Technology Review

MIT Technology Review reported on April 21–22 that Chinese open-weight AI models have surpassed US models in global download share for the first time, based on data from researchers at MIT and Hugging Face. Chinese open-weight models accounted for 17.1% of global AI model downloads over the year ending August 2025, narrowly exceeding the US share of 15.86%. The milestone reflects a strategic bet by China's leading AI labs — Alibaba, DeepSeek, Baidu, and Zhipu AI — to ship models as freely downloadable 'open-weight' packages that let developers adapt and run them without US platform gatekeepers. The strategy went mainstream after DeepSeek open-sourced its R1 reasoning model in January 2025. Alibaba's Qwen model family now has more user-generated variants on Hugging Face than models from Google and Meta combined — a structural indicator of adoption depth. Much of the Global South is embracing Chinese open-weight models as a path to 'AI sovereignty': Singapore's government-backed AI Singapore program chose Alibaba's Qwen over Meta's Llama, and Malaysia announced its sovereign AI ecosystem would run on DeepSeek. The finding represents a major inflection point — China's open-source strategy is winning the global AI distribution battle in developing markets where US AI platforms face access restrictions, cost barriers, or sovereignty concerns.