Alibaba has opened a preview of Qwen4-Max, a flagship model from its coming Qwen4 generation. The preview went live at 10:00 on November 11, 2026, Beijing time (6:00 PM PT on November 10), and it has a fixed end date. Alibaba Cloud's documentation says the model will be retired when the preview closes on November 25.
That makes this a test window, not a full Qwen4 launch. Below is what Alibaba has published, how to try the model, what it costs, and what is still missing.
What is Qwen4-Max Preview?
Qwen's official account announced it on X at 02:01 UTC on November 11 with "Qwen4 is coming." The post promised "Deeper knowledge. Richer world understanding. Stronger coding and agentic capabilities." It said Qwen4 is built for real-world tasks, from complex coding and knowledge work to front-end builds and professional software workflows. Those are Qwen's own claims, not measured results.
According to Alibaba Cloud Model Studio's page, Qwen4-Max-Preview is "the next-generation flagship multimodal model in the Qwen family." The page says it combines text generation, image and video understanding, and deep reasoning, and that it is designed for coding, office productivity, frontend design and complex agentic tasks.
The same page lists these features:
- Web search, function calling and built-in tools
- Structured outputs in JSON Object and JSON Schema formats
- Compatibility with the OpenAI Chat Completions, OpenAI Responses and Anthropic Messages APIs
- Model ID
qwen4-max-preview
How big is the context window?
Alibaba Cloud's documentation lists a context length of 1M tokens and a maximum output of 393,216 tokens. That output budget is shared by the model's thinking and its final answer.
Thinking is on by default. Developers can set reasoning_effort to low, medium or xhigh, and xhigh is the default, per the same page. The page also flags one change that can break existing code: for OpenAI Chat Completions, starting with this model, max_tokens equals max_completion_tokens and covers thinking plus the answer. Alibaba warns that a limit sized only for the answer may cut output short.
How can you try it?
Alibaba Cloud's page says the preview is available through Token Plan only, Alibaba's subscription for AI coding and agent tools. During Double 11, Token Plan monthly cards also sold on Tmall from 50% off. Pay-as-you-go access is not supported. The preview runs from 10:00 on November 11 to 24:00 on November 25, 2026 (UTC+8), which is 8:00 AM PT on November 25. Alibaba says it will make no changes to the model during that period. Qwen's announcement on X also lists early access on QwenWork, Qwen APP and QwenStudio, besides Qoder.
The model is also in Qoder, Alibaba's AI coding tool. Qoder said on X that "Qoduck has earned its name." According to Qoder, Qoduck was a Qwen test build that Qoder users ran for 4 anonymous days, updated daily, and it now upgrades to Qwen4-Max-Preview. Qoder says the model is live on every Qoder product and every plan, Teams and Enterprise included.
The same day, Alibaba opened a preview of Qwen Code 1.0, its open-source terminal coding agent, which now includes the agent harness built for Qoder.
How much does it cost?
QwenCloud's Token Plan Individual page lists these monthly plans as of November 11, 2026:
- Lite: $6 per month (limited-time new-user price $3), 20,000 Credits
- Essential: $10 per month (limited-time new-user price $5), 45,000 Credits
- Standard: $18 per month, 80,000 Credits
- Pro: $43 per month, 200,000 Credits
The same page says that during a limited-time promotion, Qwen4-Max-Preview uses Credits at as low as 20% of the standard rate, which it calls 5x usage. Calls between 22:00 and 08:00 (UTC+8) get an additional 60% off, which the page says is 8% of the standard rate. The page does not say how many Credits one Qwen4-Max request uses, so these numbers do not tell you how much work a plan covers.
Qoder lists its own offer for a limited time: "80% off, 0.1x credits," per its post.
What has Alibaba not published?
On the two documentation pages above, we found no benchmark scores, no parameter count, no technical report, no download of model weights and no pay-as-you-go price, as checked on November 11, 2026. Alibaba has also not said when the full Qwen4 release will come or which other Qwen4 models will follow.
How this compares with the Qwen4 timing reports
On October 29 we covered what Alibaba said in Amsterdam, where Qwen4 was reported to be coming the next week, a reported November 2 to 8 window. The preview opened on November 11 Beijing time, three days after that window closed, and it covers one model in preview, not a Qwen4 family.
For background on the model line, see our explainers on what Qwen is and on Qwen3.8, the generation before it.
Why it matters
The fixed window shapes what a test can tell you. Alibaba says the model will not change during the preview and will be retired at the end, so results from these two weeks describe this preview build, not a final Qwen4-Max. Developers who want to test long documents or multi-step agent work can do it now inside Token Plan or Qoder. Anyone deciding whether Qwen4 will be open-weight, or how it scores against other frontier models, still has to wait for Alibaba to publish weights, a model card or benchmark results.
