Alibaba Group’s open-weight Qwen family of artificial intelligence models has surpassed 3 billion global downloads in the past six months, overtaking competitors from Google and Meta to become the world’s most downloaded open-source AI model family, the company said.
According to data cited from Hugging Face, a leading open-source AI platform, Google’s models recorded about 418 million downloads while Meta’s stood at 227 million during the same period in 2026. Alibaba said Qwen now accounts for more than half of global open-source model downloads.The Chinese technology company has open-sourced more than 460 models under the Qwen brand.
Its ecosystem has generated over 300,000 derivative models created by developers fine-tuning or building upon the original releases.Hugging Face described Qwen as “one of the largest foundations of the open AI ecosystem,” noting that it has become part of the default workflow for many developers deciding which models to fine-tune and deploy.
The milestone underscores the rapid global adoption of Chinese open-weight models despite ongoing U.S. export controls on advanced chips and AI systems. Alibaba has been steadily expanding Qwen’s capabilities, including recent releases with vision features, longer context windows and improved performance on coding and agent benchmarks.Open-source AI models allow developers and companies to download, modify and run the systems on their own hardware without relying solely on proprietary cloud services.
This approach has gained traction among organizations seeking lower costs, greater customization and reduced dependence on a handful of large U.S. technology firms.
Alibaba’s cloud and AI division has positioned Qwen as a key part of its strategy to compete in the global generative AI market. The company continues to release updated versions at a rapid pace, contributing to the model’s growing popularity among researchers, startups and enterprises.The download figures highlight a broader shift in the open-source AI landscape, where Chinese models have closed performance gaps with leading Western systems while offering permissive licenses that encourage widespread reuse and adaptation.
