Alibaba Cloud and the Qwen team, makers of the Qwen models

Who makes Qwen?

Alibaba Cloud, founded in 2009, is the cloud computing and artificial intelligence division of Alibaba Group, based in Hangzhou, China. Inside it, the Qwen team develops the family of large language models of the same name. The models power Alibaba’s own services, including the Quark app, and are offered to the company’s cloud customers.

A largely open strategy

What made Qwen known outside China is its release policy. Alibaba publishes many open-weight models in every size, from small models that run on a laptop to very large mixture-of-experts models, an architecture that activates only part of the network for each word. Most are under Apache 2.0.

With the Qwen3 generation, Alibaba introduced “hybrid” models that can switch between a deep-thinking mode, useful for math or code, and a fast-answer mode. The lineup also includes variants for coding, vision, text-to-speech and other uses.

Not everything is open, though. The Qwen-Max, Qwen-Plus and Qwen-Flash offerings are served through Alibaba Cloud’s Model Studio API and billed by usage. That is why we list Alibaba among the “mixed” labs.

How to access Qwen models

• Qwen Chat, the online assistant: see the Qwen Chat page.
• Model Studio, Alibaba Cloud’s API platform, compatible with the OpenAI format.
• Downloads from Hugging Face, ModelScope or GitHub, to host the model yourself or fine-tune it.

What sets Alibaba apart

In practice, Qwen is often picked as the base for open projects, thanks to its range of sizes and its permissive license. The main thing to check is hosting: calling a Chinese company’s API and running a downloaded model on your own servers do not send your data down the same path. To compare it with DeepSeek or Mistral, the open source models page lists the most widely used open models.