In the spring of 2026, the open source AI community on Hugging Face no longer resembles a niche forum. It has become a global industrial workshop. The platform now hosts over 2 million public models and half a million datasets, serving 11 million users. But the raw numbers tell only part of the story. The real shift is in who is building, and where.
A fundamental geographic rebalancing is underway. Data from the past year shows China has overtaken the United States in monthly model downloads, accounting for 41% of all downloads. This surge followed the viral release of DeepSeek's R1 model in early 2025, which prompted a strategic pivot. Chinese tech giants like Baidu, which had zero releases on the Hub in 2024, published over 100 models in 2025. ByteDance and Tencent increased their output eight to ninefold.
Meanwhile, the composition of developers has flipped. Independent, unaffiliated creators now account for 39% of all downloads, nearly matching industry's share, which fell to 37%. Individuals and small collectives specializing in quantizing and adapting base models now steer a significant portion of what typical users can run.
The action is moving to specialized, practical domains. Robotics datasets exploded from just over 1,100 in 2024 to nearly 27,000 in 2025, making it the largest dataset category on the Hub. Scientific research for drug discovery and protein folding is another hot zone, with community-led projects coordinating hundreds of contributors across disciplines.
This activity is increasingly tied to national strategy. South Korea's Sovereign AI Initiative named national champions to produce domestic models, with three trending simultaneously on the Hub in February. The UK's 'public money, public code' principle and EU-funded projects reflect similar priorities, using open models to ensure deployment on domestic hardware under local legal frameworks.
The message from the data is clear: open source AI is no longer a sideshow. It's the primary arena where practical AI gets built, adapted, and deployed for the real world.
Source: Hugging Face Blog