Alibaba’s Qwen tops 3bn AI model downloads

Alibaba’s (BABA) Qwen family of open-weight AI models has accumulated more than 3 billion global downloads over the past six months. These downloads reflect how often developers and organizations have obtained the models to run or adapt them for their own applications.

Open-weight models such as Qwen can be downloaded, customized and reused as building blocks for new AI products. Download volumes are increasingly used as a proxy for how widely a model family is being tested, integrated and experimented with by the developer community.

A state-of-open-models report released on August 14 shows that in 2026, open models from Google reached 418 million downloads, while Meta’s models reached 227 million. The latest six‑month tally for Qwen therefore places Alibaba’s open-weight family ahead of these two U.S. peers on raw download counts.

The same analysis characterizes Qwen as one of the largest foundations in the open AI ecosystem. In this context, Qwen’s scale is assessed not only by downloads but also by the breadth of model variants and the activity around derivative projects.

Alibaba has open-sourced more than 460 models within the Qwen family. These include multiple configurations that target different performance, size or use-case trade-offs for developers.

Around this core, the Qwen ecosystem has spawned over 300,000 derivative models. These derivatives are created when developers fine-tune, adapt or otherwise modify base Qwen models to address specific domains, languages or tasks.

High derivative counts suggest that Qwen is being actively used as a foundation for downstream projects. This activity helps entrench the models in workflows and can encourage further experimentation and tooling built around the Qwen stack.

Alibaba distributes Qwen through its cloud platform to enterprise customers in multiple regions. Markets highlighted for this distribution include Southeast Asia and Africa, extending Qwen’s reach beyond its domestic base.

Cloud delivery lowers the barrier for enterprises to access and deploy Qwen models, supporting use in production systems as well as in testing environments. This channel likely contributes to both download volumes and derivative model creation as customers tailor Qwen to their needs.

The combination of extensive open-sourcing, strong derivative activity and broad cloud distribution positions Qwen as a central player in the open-model landscape. Other Chinese model builders and U.S. firms are also expanding their open offerings, indicating an intensifying competition to attract developers to their ecosystems.