State of Open Source on Hugging Face: Spring 2026

The center of gravity in the open-source AI world has shifted, with China surpassing the United States in both monthly and overall model downloads on the Hugging Face platform. According to a new analysis of platform activity, models from China now account for 41% of all downloads, a stark realignment of the global landscape. This change is not just geographic; it's also structural, as independent developers now rival industry labs in their share of total model usage, rising from 17% to 39% since 2022. The data indicates that open-source AI is evolving from a centralized, U.S.-led field into a more distributed and competitive global ecosystem.

This rapid ascent was ignited by the viral release of DeepSeek's R1 model in January 2025, which prompted a decisive strategic pivot by Chinese technology giants. Companies that previously favored closed systems, including Baidu, ByteDance, and Tencent, dramatically increased their open-source contributions, with Baidu going from zero releases in 2024 to over 100 in 2025. This surge in releases is complemented by a technical trend toward smaller, more accessible models that see higher adoption rates due to practical constraints on cost and hardware. For example, Alibaba's Qwen model family has spawned over 113,000 derivative models, showcasing how usability drives community engagement and adaptation more than scale alone.

The implications of this shift extend to national policy, as governments increasingly view open-source models as tools for digital sovereignty. Initiatives in South Korea, Switzerland, and the UK aim to build domestic AI capabilities by fine-tuning open-weight models on local data, reducing reliance on foreign-controlled infrastructure. This trend suggests a future where regional ecosystems flourish, developing models tailored to local languages and needs. The proliferation of derivative models and the rise of independent contributors show that the value of open-source AI is increasingly realized not just in the creation of large base models, but in their widespread adaptation and specialization across the globe.

The open-source AI landscape is no longer defined by a few large American labs, but by a multi-polar competition where national strategic interests and developer accessibility are becoming the primary drivers of influence and adoption.