Alibaba has opened Qwen4-Max — the first preview from its next flagship Qwen4 model family — to developers through its Token Plan subscription and in Qwen Chat, but only for a fixed window: the preview model is retired after November 25. Why it matters: it is a first real look at Alibaba’s next flagship, offered as a two-week evaluation rather than a product you can build on.
November 11, 2026
What happened
The preview, under the model ID qwen4-max-preview, began at 10:00 a.m. China time on November 11. Alibaba says the model will remain unchanged throughout the preview period and will be retired when the window closes at midnight Beijing time after November 25. Early access is also offered through QwenWork, Qoder, the Qwen app and QwenStudio, and the model has appeared in the Qwen Chat model selector as well.
According to Alibaba’s documentation, the model covers text generation and reasoning, image and video understanding, tool calls and structured outputs, aimed at coding, office work and agent tasks.
The access rules are the story
On the API side, Qwen4-Max Preview is Token Plan only: pay-as-you-go access is explicitly not supported. Token Plan uses its own dedicated key and endpoint, separate from ordinary Model Studio or Coding Plan credentials, and offers OpenAI-compatible and Anthropic-compatible protocols. The Personal Edition terms limit the plan to interactive use inside compatible coding and agent tools — application backends and automated batch jobs are outside the scope.
In plain terms: you can evaluate this model inside a coding tool for two weeks. You cannot put it behind your product, and anything you measure describes a version that stops existing on November 25.
What Alibaba has not published
No benchmarks. No model card. No parameter count. No per-token price, and no statement about open weights. Alibaba has said the Qwen4 generation improves on Qwen3.8 for world knowledge, code generation and sustained performance on long tasks with repeated tool calls — but as of today, none of that is backed by published numbers anyone can check. Anything you read quoting specifications for Qwen4-Max beyond the access and capability details above is provisional.
Why it matters
A flagship preview with an expiry date is a new kind of launch: part capability demonstration, part deadline. For developers, the honest takeaway is narrow — if you already work in Alibaba’s Token Plan tools, two weeks of free-form evaluation costs you little and tells you how the next Qwen flagship feels on your own tasks. For everyone watching the model race, the real news will be what Alibaba publishes when the full Qwen4 family ships: the benchmarks, the pricing and whether any of it arrives as open weights. None of that is on the table yet.
Frequently asked questions
How long is Qwen4-Max Preview available?
From November 11 until midnight Beijing time after November 25, 2026. Alibaba says the model stays unchanged during the window and is retired afterward.
Can I use Qwen4-Max Preview in my app?
Not through pay-as-you-go API access — Alibaba explicitly excludes it. Token Plan access is limited to interactive use in compatible coding and agent tools.
How good is Qwen4-Max?
Nobody outside Alibaba can say yet. No benchmarks, model card or parameter count have been published, so capability claims are unverified for now.
Sources: Alibaba Qwen documentation via S3DAT News, OrcaRouter, Dev.to.