Alibaba says Qwen4 is coming, with a Max preview live on Token Plan

Alibaba announced on November 11 that its next-generation Qwen model, Qwen4, is about to be released, with a preview version already live on the company's Token Plan. Chinese outlets including Sohu and Phoenix reported that the preview is Qwen4-Max and that it has been connected to several Alibaba products.

The reports did not disclose Qwen4's parameter count, context length, pricing or open-source plans. What is confirmed is narrower: the model is officially positioned as the new generation of Qwen, and the preview landed first on Token Plan, a usage-based calling service, rather than in a consumer-facing release.

Putting a preview on Token Plan means developers and enterprise customers can test the new model against real traffic before the formal launch. That release rhythm, ship to usage first and announce second, has become common among leading model makers: the official launch becomes a confirmation of specs and price rather than the first reveal of capability.

For Alibaba, the Qwen family is the core asset of its AI strategy, used to build a developer ecosystem through open versions and to monetize through cloud services and Token Plan. Placing the newest generation's preview on a metered platform ties model iteration more tightly to the commercial path.

The timing matters too. Through early November, Chinese technical communities discussed Qwen4 intensively, circulating benchmark comparisons and architectural guesses that Alibaba has not confirmed. Officially, only the coming release and the live preview are established; everything else waits on the launch.

Three things to watch: the formal release window, whether Alibaba keeps its open-source approach and which sizes it opens, and whether Token Plan pricing shifts meaningfully from the previous generation. Together they decide whether Qwen4 is a routine upgrade or a deliberate push in the enterprise AI market.