Alibaba opens Qwen4-Max preview to Token Plan subscribers through November 25

Alibaba opened a preview of Qwen4-Max on November 11, giving developers a first chance to try its next flagship model in coding tools. API access requires an Alibaba Cloud Token Plan subscription. The company says the preview model will be retired at midnight at the end of November 25 in China, or 16:00 UTC that day.

The Qwen team also announced early access in the Qwen app, Qoder, QwenWork and QwenStudio. The November 11 release notes for Alibaba Cloud Model Studio list the preview under Singapore and mark the API as exclusive to Token Plan. Qwen's announcement did not specify eligibility or an end date for each of the four apps.

Qoder's model selector documentation already lists the preview as a selectable model. It shows an estimated credit multiplier of 0.5× for Qwen4-Max Preview, the same as Qwen3.8-Max. Actual charges depend on the task and settings. That is Qoder's own rate, separate from the API subscription promotion.

What fits in the model's context

Qwen4-Max Preview accepts text, images and video and responds with text. QwenCloud lists a one-million-token context window, the amount of information the model can consider in a request. A developer can, for example, submit a long description of a workflow along with still images and a clip, then ask for a written account of what happens.

The output limit is 393,216 tokens, counting both the model's reasoning and its answer. Alibaba applies that combined budget to the output limit in its OpenAI-compatible chat interface as well. A limit previously sized for the answer alone can cut the response short when reasoning takes up part of it. The previous Qwen3.8-Max already accepted the same input types and offered the same context length. The specification lists a higher output ceiling for the preview.

The Model Studio visual-input table sets a limit of 16 million pixels per image and two hours and 2 GB per video. It lists maximum counts of 2,048 images and 64 videos, subject to the request's overall context limit. The model can also return data in a specified format and request a function supplied by the application.

Subscription fees and usage credits

The Personal Edition's Lite plan costs $6 a month and includes 20,000 credits. Essential costs $10 and includes 45,000. Alibaba lists promotional prices of $3 and $5 for accounts that have never subscribed to Personal Edition, available until just before midnight on November 30 in China. The number of credits deducted for each request depends on token use, the model, reasoning mode and tool calls.

A separate launch promotion affects the credits used by Qwen4-Max Preview. QwenCloud says eligible calls can use as little as 20% of the standard credits, falling to 8% between 22:00 and 08:00 China time. These figures describe usage inside a subscription. The introductory prices for the subscription itself follow their own schedule.

Alibaba describes Personal Edition as intended for interactive use in coding and agent tools. Its documentation excludes automation scripts, custom application backends and non-interactive batch calls. The plan is offered in the Singapore region with global model deployment, which Alibaba says can involve cross-border transfers of prompts and outputs.

The API model is listed as qwen4-max-preview. Alibaba says the version will stay fixed for the preview period, from 10:00 on November 11 through the end of November 25 in China, and will be retired when the window closes. The announcement gives no date for a general release or successor to this preview.