Alibaba has launched Qwen4-Max, the new flagship of its Qwen family of AI models, and is keeping the model's weights closed.
The Qwen team announced the model on Monday, November 16, in a launch blog post. The post pitches Qwen4-Max as an AI teammate for work, creativity and discovery. That is the company's own framing, and the details we have do not include benchmark scores, pricing or model size.
"Weights" are the trained numbers that make up a model. When a lab publishes them, anyone can download the model, run it on their own hardware, inspect it and adapt it. When a lab withholds them, the model stays under the lab's control, and outsiders can use only what the lab chooses to offer.
The "teammate" pitch fits that choice. It sells a service, not a file, and a service is easier to charge for and easier to change after launch.
The split is familiar across the industry. OpenAI and Anthropic keep the weights of their top models closed, while Meta's Llama models are the best-known example of releasing them. Qwen4-Max puts Alibaba's flagship on the closed side of that line.
What Alibaba will charge, and where developers can reach the model, is the next thing to look for.