Skip to content
NLEN
Illustration: a large AI model shown as a glass building with an hourglass in front and a calendar with a circled end date.

Qwen4-Max-Preview: Alibaba's new top model, Token Plan only and gone after November 25

By Ivo Donker · — text: Claude (AI, Anthropic), checked against the sources

Can you use Qwen4-Max-Preview, and for how long? Yes, but only through the Token Plan of Alibaba Cloud Model Studio and only temporarily. The preview runs from November 11, 2026, 10:00, to November 25, 2026, 24:00 (China time, UTC+8). Alibaba itself writes that the model will be retired after that. Pay-as-you-go access is not supported.

We read Alibaba Cloud's documentation page, updated on November 11, 2026. The topic was also in the trending briefing of November 12.

What the model can do

The model ID is qwen4-max-preview. The context window is 1 million tokens. The model generates text, understands images and video, and supports function calling, web search, built-in tools and structured output (JSON Object and JSON Schema). It can be called through the OpenAI Chat Completions and Responses APIs and through the Anthropic Messages API. You get an API key through Token Plan Personal or Token Plan Team.

The pitfall: thinking and answer share one budget

Maximum output is 393,216 tokens, and thinking and the final answer share that budget. Thinking is on by default, at the highest level xhigh; you can also choose low or medium. Alibaba warns: "Thinking consumes part of the same budget, so a limit previously sized for the answer alone may truncate output." Anyone with an agent or script whose max_tokens limit was sized for the answer alone may get truncated answers. More on how thinking models work in reasoning models.

An end date is part of the design

Alibaba writes that the model will not change during the preview and "will be retired" after it. A price per token is not on the page; it refers to the Token Plan guides. The page also does not mention benchmarks, parameter count or open weights. That makes it a trial model, not a fixed choice for anything that must keep working after November 25. The broader trade-off between open and closed models is in open versus closed models.

What we did not check

We did not test the model ourselves and did not read the Token Plan prices. There are no benchmarks on the page we read; we did not use scores circulating on social media.

Checked on November 12, 2026.

This article was written with AI (Claude, Anthropic) and checked against the sources below.

Sources