Last updated: November 11, 2026
Qwen4-Max-Preview is Alibaba’s preview of its next flagship Qwen model, released November 11, 2026. It reads text, images and video, reasons by default and has a 1M-token context. Developers can use it only through a Token Plan subscription until November 25, 2026, after which the preview model is retired.
Contents
- What Alibaba released
- Who can use it and what it costs
- How to set up Qwen4-Max-Preview
- API settings power users should know
- Qwen4-Max-Preview vs qwen3.8-max
- Limitations
- Frequently asked questions
This guide explains what the preview includes, how to reach it, and which settings change compared with earlier Qwen models. Facts come from Alibaba Cloud’s Qwen4-Max-Preview model page, the Token Plan Personal Edition overview and related Model Studio documentation.
What Alibaba released
Alibaba Cloud describes Qwen4-Max-Preview as “the next-generation flagship multimodal model in the Qwen family.” It is designed for coding, office productivity, frontend design and complex agent tasks. According to TechNode, the preview opened on Wednesday, November 11, and early access is also offered in the Qwen app, Qoder, QwenWork and QwenStudio.
| Item | Details |
|---|---|
| Model ID | qwen4-max-preview (alias qwen4.0-max-preview) |
| Input | Text, images and video |
| Context length | 1M tokens |
| Maximum output | 393,216 tokens, shared by thinking and the final answer |
| Reasoning | Thinking on by default; reasoning_effort low, medium or xhigh (default) |
| Tools | Function calling, web search and built-in tools |
| Structured outputs | JSON Object and JSON Schema |
| API protocols | OpenAI Chat Completions, OpenAI Responses and Anthropic Messages |
| Preview window | Nov 11, 2026 10:00 to Nov 25, 2026 24:00 (UTC+8) |
| Access | Token Plan only; no pay-as-you-go |
Alibaba has not published a parameter count, benchmark scores or a pay-as-you-go token price on these pages. Treat performance claims from third parties with caution until official numbers appear.
Who can use it and what it costs
Developers need Token Plan, a Model Studio subscription billed in Credits. Personal Edition is available only in the Singapore region. Usage is deducted from the monthly Credits quota, and the amount per request depends on the model, tokens used, thinking mode and tool calls.
| Personal plan | Price per month | Monthly Credits | Concurrent agents |
|---|---|---|---|
| Lite | $6 ($3 new-user price) | 20,000 | 1-2 |
| Essential | $10 ($5 new-user price) | 45,000 | 2-3 |
| Standard | $18 | 80,000 | 3-4 |
| Pro | $43 | 200,000 | 6-8 |
The new-user prices run until 23:59:59 on November 30, 2026 (UTC+8) for accounts that have never bought Personal Edition. Extra bundles cost $15 per month for 20,000 Credits, and you can hold up to five.
The preview carries a promotion: Qwen4-Max-Preview uses Credits at as little as 20% of the standard rate. Calls between 22:00 and 08:00 (UTC+8) get a further 60% off, or 8% of the standard rate. That window is 9:00 a.m. to 7:00 p.m. Eastern Time while the U.S. is on standard time, which suits heavy testing during a U.S. workday.
For other ways to use Qwen at no cost, see our answer on whether Qwen is free.
How to set up Qwen4-Max-Preview
- Subscribe to Token Plan Personal Edition (or get a seat on a Team Edition plan) in the Model Studio console.
- Generate an API key on the My Subscription page. Alibaba shows it in full only once, so store it securely.
- Pick the base URL for your tool’s protocol: https://token-plan.ap-southeast-1.maas.aliyuncs.com/compatible-mode/v1 for OpenAI-compatible tools, or https://token-plan.ap-southeast-1.maas.aliyuncs.com/apps/anthropic for Anthropic-compatible tools.
- Set the model to qwen4-max-preview in your tool.
- Send a short test prompt and check usage in the console’s subscription usage details.
For Claude Code, Alibaba’s Claude Code guide sets ANTHROPIC_BASE_URL to the /apps/anthropic endpoint and maps ANTHROPIC_DEFAULT_OPUS_MODEL to qwen4-max-preview, with CLAUDE_CODE_MAX_CONTEXT_TOKENS set to 983616. If long output is truncated, Alibaba suggests setting CLAUDE_CODE_MAX_OUTPUT_TOKENS=64000. A 401 invalid_api_key error usually means the key type does not match the base URL. Our Claude Code 2.1.318 guide covers recent Claude Code changes.
The Token Plan quick start lists other supported tools, including Cursor, Codex, Qwen Code, OpenCode, Cline, Qoder, Cherry Studio and Chatbox. For Alibaba’s own coding agent, see our Qwen Code 0.26 guide, and for the desktop office agent, our QwenWork 2.0 guide.
API settings power users should know
max_tokens now includes thinking. Alibaba’s API compatibility notice says that starting with qwen4-max-preview, max_tokens in Chat Completions equals max_completion_tokens and limits thinking plus the final answer. A limit sized for the answer alone may cut off output. Older Qwen models keep their previous behavior.
Reasoning effort. Thinking is on by default at xhigh. Use reasoning_effort low or medium for faster, cheaper replies. Do not set reasoning_effort and thinking_budget together.
Multi-turn thinking. Chat Completions turns on preserve_thinking by default. Return earlier thinking in the reasoning_content field of assistant messages, not inside content.
Built-in tools. Through the Responses API, Token Plan offers web_search, web_search_image, image_search, web_extractor and code_interpreter, with Credits deducted per use.
Qwen4-Max-Preview vs qwen3.8-max
Both models appear in the Token Plan model list with reasoning, visual understanding and text generation. The practical differences are access and behavior. Qwen4-Max-Preview is documented with text, image and video input, runs only for a fixed preview window, uses the new combined max_tokens budget, and gets the 20% Credits promotion plus the 8% night rate. qwen3.8-max stays available after November 25 and has its own 60% night discount. Keep qwen3.8-max as a fallback in any tool configuration you build during the preview.
Limitations
- The preview model is retired after November 25, 2026, so do not build production work on it.
- Token Plan Personal Edition is for interactive use in coding and agent tools only. Automation scripts, custom application backends and batch jobs are not allowed.
- Inference runs in a global deployment from the Singapore region, so prompts and outputs involve cross-border data transfer.
- Personal Edition cannot currently be unsubscribed, and keys may not be shared with others.
- No official benchmarks are published yet. Compare it with models such as ERNIE 5.5 Flash Preview on your own tasks.
Frequently asked questions
What is Qwen4-Max-Preview?
Qwen4-Max-Preview is a preview of the next-generation flagship model in Alibaba’s Qwen family. It combines text generation, image and video understanding and deep reasoning, with a 1M-token context window, and is aimed at coding, office work, frontend design and agentic tasks.
How long is the Qwen4-Max-Preview available?
Alibaba Cloud lists the preview period as 10:00 on November 11, 2026 to 24:00 on November 25, 2026 (UTC+8). The model will not change during that period and will be retired when the preview ends.
Can I use Qwen4-Max-Preview with pay-as-you-go API billing?
No. Alibaba Cloud says the preview is available through Token Plan only, and pay-as-you-go access is not supported. You need a Token Plan Personal or Team subscription and its dedicated API key.
How much does it cost to try Qwen4-Max-Preview?
There is no separate model price. Usage is deducted from Token Plan Credits. Token Plan Personal Edition costs $6, $10, $18 or $43 per month, with new-user prices of $3 and $5 on the two lower tiers until November 30, 2026 (UTC+8). During the preview, the model uses Credits at as little as 20% of the standard rate.
What is the model ID for Qwen4-Max-Preview?
The model ID is qwen4-max-preview. Alibaba Cloud also accepts the alias qwen4.0-max-preview, and bills show the alias name.
Can I use Qwen4-Max-Preview in an application backend?
Not under Token Plan Personal Edition terms. The plan is for interactive use in coding and agent tools such as Claude Code, Cursor, Qwen Code, Qoder and OpenClaw. Automation scripts, custom application backends and batch calls are not allowed and may lead to suspension.
Sources
- qwen4-max-preview, Alibaba Cloud Model Studio documentation
- Token Plan Personal Edition overview, Alibaba Cloud Model Studio
- Token Plan Personal Edition quick start, Alibaba Cloud Model Studio
- API compatibility upgrades, Alibaba Cloud Model Studio
- Alibaba opens Qwen4-Max preview with a million-token context window, TechNode (November 11, 2026)