Qwen4-Max preview: specs, access and how to try it.

Qwen4-Max is a model from Alibaba’s upcoming Qwen4 family, and it went into preview on 11 November 2026. The Qwen team says Qwen4 brings “deeper knowledge, richer world understanding” and “stronger coding and agentic capabilities”, built for complex coding, knowledge work, front-end builds and professional software workflows. Here is what the Qwen4-Max preview is, what the docs confirm about its specs, how access works and how to try it before the preview ends on 25 November.

What is Qwen4-Max?

Alibaba Cloud’s Model Studio docs describe Qwen4-Max-Preview as “the next-generation flagship multimodal model in the Qwen family”. It combines text generation, image and video understanding and deep reasoning, and Alibaba says it is designed for coding, office productivity, frontend design and complex agentic tasks. The previous Max model on The Model Atlas is Qwen3.8-Max.

The model ID is `qwen4-max-preview`. QwenCloud’s Token Plan docs add that `qwen4.0-max-preview` is an alias: both IDs work, and bills show `qwen4.0-max-preview`.

This is a preview, not a full release. Alibaba says the preview runs from 10:00 on 11 November 2026 to 24:00 on 25 November 2026 (UTC+8), that no changes will be made to the model during that period, and that the model will be retired when the preview ends. Alibaba has not published benchmark scores, a parameter count or open weights for this preview.

Qwen4-Max preview specs and features

According to the Model Studio docs, the Qwen4-Max preview supports:

• 1M-token context and up to 393,216 output tokens, shared between thinking and the final answer.
• Deep reasoning on by default. Thinking is enabled by default, and `reasoning_effort` can be set to `low`, `medium` or `xhigh` (the default). Alibaba says not to set `reasoning_effort` and `thinking_budget` together.
• Image and video input. It accepts text, images and video for tasks like screenshot analysis, document understanding and video analysis.
• Tool calling. Function calling, web search and built-in tools.
• Structured outputs in JSON Object and JSON Schema modes.
• API compatibility with the OpenAI Chat Completions and Responses APIs and the Anthropic Messages API.

One change matters if you move existing code over. Alibaba says that starting with `qwen4-max-preview`, `max_tokens` is equivalent to `max_completion_tokens` in OpenAI Chat Completions, so the limit covers thinking and the answer together. A limit sized for the answer alone may cut output short. Chat Completions also turns on `preserve_thinking` by default for multi-turn chats; Alibaba asks developers to return earlier thinking in the `reasoning_content` field rather than in `content`.

Qwen4-Max price and Token Plan access

Qwen4-Max has no pay-as-you-go price during the preview. Alibaba says it is available through Token Plan only, its subscription plan for coding and agent tools. Token Plan bills in Credits, and QwenCloud says the Credits a request uses depend on the model, token usage, thinking mode and tool calls.

QwenCloud’s Token Plan docs list the Individual plans as Lite at $6 a month (20,000 Credits), Essential at $10 (45,000 Credits), Standard at $18 (80,000 Credits) and Pro at $43 (200,000 Credits), with limited-time new-user prices of $3 for Lite and $5 for Essential. For the Qwen4-Max preview, QwenCloud says Credits use is as low as 20% of the standard rate (“5x usage”) during a limited-time promotion, and calls between 22:00 and 08:00 (UTC+8) get an extra 60% off, or 8% of the standard rate. QwenCloud notes it may change promotions.

Token Plan has usage rules worth reading first. QwenCloud says it is limited to interactive use inside programming and agent tools such as Claude Code, Cursor, Qwen Code, Qoder and OpenClaw, and must not be used for automation scripts, custom application backends or non-interactive batch calls.

How to try Qwen4-Max

The Qwen team’s announcement says you can get early access on QwenWork, Qoder, the Qwen app and QwenStudio, as well as through Token Plan. For API access, Alibaba’s docs give three steps:

1. Choose Token Plan Personal or Token Plan Team and follow its guide to get a dedicated API key.
2. Use the dedicated Token Plan base URL listed in the docs for your region and protocol.
3. Set the model ID to `qwen4-max-preview`.

Because the model accepts OpenAI- and Anthropic-style requests, tools that already speak those APIs can point at it by changing the base URL, key and model ID. Leave room in your output limit for thinking, since both share one budget.

What people are building with Qwen4-Max

The Qwen team asked developers to “put it to the test on a real problem”, and the first builds came in within a day. Two are on The Model Atlas so far, both made in Qoder:

• Restaurant website built with Qwen4-Max: a café site with a menu, food gallery, about section and table reservations.
• Campaign brief desk app built with Qwen4-Max (sponsored post): an app that organizes campaign briefs and checks deadlines across Beijing, US Eastern and US Pacific time, which caught a one-hour conflict in a test brief.

More builds are on the Qwen4-Max model page; others may be added after curation.

Who is Qwen4-Max for?

Going by Alibaba’s own description, the preview is aimed at developers who do agentic coding, front-end work and long, multi-step tasks inside coding tools, and who want to test a new flagship model early at a discounted Credits rate. Because it is Token Plan only, limited to interactive tool use and retired on 25 November, it is a model to evaluate, not one to build a production backend on yet.