According to the latest open-source large AI model rankings released by Hugging Face, the world’s largest AI open-source community, Alibaba’s open-source native multimodal model, Qwen 3.5, has secured the top spot. Launched on Chinese New Year’s Eve, this model boasts an impressive total of 397 billion parameters, with only 17 billion activated. Its performance is on par with that of Gemini 3, yet its token costs are merely around 5% of Gemini 3’s. Upon its release, the model swiftly drew responses from the global industrial chain, with numerous international tech companies as well as domestic GPUs and platforms successfully completing adaptations. To date, Alibaba’s Qwen has open-sourced over 400 models, with global downloads surpassing 1 billion and more than 200,000 derivative models created. Among the top 10 models in the rankings, 8 hail from Chinese teams. Over the past year, the global download share of open-source models developed in China has, for the first time, exceeded that of the United States. It is projected that by 2025, China will be home to over 6,000 AI enterprises, with the core industry scale exceeding RMB 1.2 trillion.