The global open source AI landscape is undergoing a major reshuffle, with Chinese open source models rapidly emerging.

The latest observational report released by the world's largest AI open-source platform, Hugging Face, shows that the global open-source artificial intelligence ecosystem is rapidly expanding, with Chinese models becoming an essential force within it. The report points out that for most months since 2026, the largest open-source models released by Chinese AI companies have significantly outpaced their American counterparts in parameter scale: the monthly parameter limit for Chinese models ranges from 754 billion to 2.78 trillion, while most American models during the same period were below 130 billion. Not only in scale, but Chinese models also excel in building an open ecosystem. Data from Hugging Face indicates that over 80% of Chinese models with parameters exceeding 20 billion adopt relatively permissive licensing agreements, allowing companies and developers to engage in secondary development and commercial deployment. Among these, Qwen has become an important foundational model in the Hugging Face open-source ecosystem. Data shows that the number of derivative models based on Qwen has surpassed 150,000, far exceeding competitors like Google and Meta. The Hugging Face report also notes that the two companies that have introduced the most new open-source models in the U.S. this year are no longer Google and Meta, but AMD and NVIDIA. NVIDIA has launched the Nemotron series of models, showcasing the hardware synergy of GPUs through open models; AMD is more involved in model conversion and performance tuning.