Alibaba opened a preview of Qwen4-Max, the top model of its next Qwen generation, on 11 November at 10:00 Beijing time. You can try it until the end of 25 November in the Qwen app and Alibaba’s own work and coding tools, or through a Token Plan subscription that costs $3 a month for new users. Alibaba has not published a benchmark score, a parameter count or a regular API price, and it retires the model when the preview ends.
What is Qwen4-Max good at, according to Alibaba?
Alibaba lists Qwen4-Max for complex reasoning, coding and visual understanding, but gives no numbers to back that up. TechNode and BigGo Finance, citing Alibaba’s documentation, report a context window of one million tokens. That is the amount of text the model can read in one go. They also report answers of up to 393,216 tokens, and text, images and video as input. According to BigGo Finance, Alibaba called the model’s results in practical task tests impressive, without publishing those tests.
That makes Qwen4-Max hard to place. Qwen3.8-Max, the current top model, is still listed beside it. Alibaba already called the smaller Qwen3.8-Flash “an early preview of Qwen4” and published benchmark scores for that model. For Qwen4-Max itself, the only comparison you will have is the one you run yourself.
The $3 Token Plan is meant for coding tools, not for your own applications
For developers, the way in is Token Plan Personal Edition, a monthly Alibaba Cloud subscription that you spend in credits. BigGo Finance reports that the preview cannot be used with pay-as-you-go billing. Alibaba’s terms limit the plan to interactive use in coding and agent tools such as Claude Code, Cursor, Qwen Code and Qoder. Automation scripts, application back ends and batch jobs are not allowed, and the subscription is for one person only. You also cannot cancel a Personal Edition subscription at the moment, so check the terms before you sign up.
- Model ID: qwen4-max-preview (alias qwen4.0-max-preview)
- Preview period: 11 November, 10:00, to 25 November, 24:00, Beijing time. In the Netherlands, it ends at 17:00 on 25 November.
- Price: Lite 6amonth(3 for new subscribers until 30 November), Essential 10(5), Standard $18, Pro $43
- Discount: during Alibaba’s current promotion, the preview uses credits at as little as 20% of the standard rate, and 8% between 22:00 and 08:00 Beijing time
- Region: Singapore only
- Other ways to try it: the Qwen app, Qwen Studio, QwenWork and Qoder, according to TechNode
Your prompts leave Europe, even though Alibaba Cloud now has an Amsterdam region
Token Plan Personal Edition runs only in Alibaba Cloud’s Singapore region, in what Alibaba calls Global deployment mode. Alibaba states that the model then runs on servers around the world and that prompts and answers involve cross-border data transfer. Alibaba Cloud opened its first Dutch cloud region in Amsterdam on 29 October. Neither that announcement nor the Token Plan documentation says Qwen models run there. Alibaba also does not say where the Qwen app processes chats from European users.
For organisations in the Netherlands, that decides what you can put into the preview. Personal data and client code would leave the EU without a clear destination, which you would have to justify under the GDPR. Public code, synthetic data and your own test prompts are fine. In DataNorth’s work on model choice and sovereign AI, the first question is rarely which model scores highest. It is whether the model can run where your data is allowed to go. Qwen4-Max cannot answer that yet: there is no EU region, no regular API and no model left after 25 November.
Test Qwen4-Max on public code before 25 November, and wait for API prices before you plan
Teams that compare models regularly can use these two weeks well. Pick ten to twenty tasks from your own work that contain nothing confidential, run them through Qwen4-Max and the model you use now, and save the prompts and answers. When Alibaba releases the final version, you can rerun the same set within a day. Teams that just need a model that works can wait. Alibaba freezes the preview until it ends, and nothing you build on it will run after 25 November.
For more information, visit the Qwen4-Max Preview entry in Alibaba Cloud’s Token Plan documentation.
