Alibaba opens a 15-day Qwen4-Max preview with a 1M-token context on its Token Plan
Alibaba opened a Qwen4-Max preview on 11 November 2026, its first Qwen4 model, with a 1M-token context, sold only through its Token Plan until 25 November.
By Emre A.Published
Tech AI LatestAlibaba's Qwen team opened a preview of Qwen4-Max on 11 November 2026, the first model of the Qwen4 generation, aimed at coding, knowledge work and agent tasks. The preview runs for 15 days, until 25 November, with a context window of 1 million tokens, and is available only through Alibaba's Token Plan subscription and Qwen's own apps. Alibaba has published no benchmarks, parameter count, model card or open weights.
Key facts
- Announced: 11 November 2026 (Qwen on X)
- Context window: 1 million tokens; maximum output 393,216 tokens, thinking included
- Preview window: 11 to 25 November 2026, after which the preview model is retired
- Access: Token Plan subscription (credit based, Lite tier listed at $6 a month, $3 for new users), plus QwenWork, Qoder, the Qwen app and QwenStudio
- Not stated: benchmarks, parameter count, per-token price after the preview, open weights, full release date
What Alibaba announced
The Qwen team at Alibaba announced on 11 November 2026 that Qwen4-Max is in preview on Alibaba's Token Plan. The post describes Qwen4 as bringing deeper knowledge, richer world understanding and stronger coding and agentic capabilities, built for real-world tasks from complex coding and knowledge work to front-end builds and professional software workflows. Early access is also open in Alibaba's own products: QwenWork, the Qoder coding tool, the Qwen app and QwenStudio.
The preview has a fixed life. Alibaba Cloud's model listing, as reported by several outlets that checked it, puts the end of the preview at 25 November 2026, fifteen days after launch, and says the preview model is retired at that point. The context window is 1 million tokens, and the maximum output is 393,216 tokens, a limit that covers both the model's thinking and its answer.
This is the first model of the Qwen4 generation. Alibaba has not said when the full Qwen4-Max release will follow, or whether smaller Qwen4 models will be published with open weights as many Qwen3 models were.
How access and pricing work
Qwen4-Max is not on the regular pay-per-token price list of Alibaba Cloud Model Studio. It is reachable only through the Token Plan, a subscription paid in credits that needs its own Token Plan key and, according to developer write-ups, a Singapore endpoint rather than a standard Model Studio key. Coverage of the plan lists a Lite tier at $6 a month for 20,000 credits, with a $3 introductory price for new users.
Because usage is billed in credits rather than in dollars per million tokens, there is no published per-token price that can be set beside GPT, Claude or Gemini rates, and the price after the preview is unknown. Teams that build on the preview should expect the model name, limits and pricing to change when the full release arrives.
What is missing and why it matters
Alibaba has published no benchmark table, model card, parameter count, licence or downloadable weights for Qwen4-Max. Claims circulating on social media that it beats other frontier models are therefore not supported by anything from Alibaba. That makes this preview a hands-on trial for developers rather than a measurable release, and any comparison has to wait for either Alibaba's numbers or independent evaluations.
The release still matters for the AI model market. Qwen is one of the most widely used model families among developers who self-host, and a new flagship sets expectations for the open-weight models that usually follow it. It also arrives weeks after Cohere reported unexpected network behaviour from a Qwen3.8 model in a sandbox, so enterprises outside China are likely to ask for a model card and safety documentation before adopting Qwen4 in production. The dates to watch are 25 November, when the preview closes, and the first published benchmarks.
What changed
| Before | Alibaba's newest flagship family was Qwen3.8; no Qwen4 model was available |
|---|---|
| After | Qwen4-Max is in a 15-day preview with a 1M-token context, available only through the Token Plan and Qwen apps |
| Who it affects | Developers on Alibaba Cloud and Qwen apps; teams comparing frontier coding models; users of Qwen open-weight models awaiting Qwen4 |
Questions people ask
- How long is the Qwen4-Max preview available?
- From 11 November to 25 November 2026. The preview model is retired when the preview closes, according to Alibaba Cloud's model listing as reported by several outlets.
- How much does Qwen4-Max cost?
- There is no per-token price. Access is through Alibaba's credit-based Token Plan, whose Lite tier is listed at $6 a month, or $3 for new users.
- Are Qwen4-Max weights open?
- No. Alibaba has not released weights, a licence, a model card or benchmarks for the preview.
Sources
- Qwen4 is coming: Qwen4-Max is now in preview on Alibaba's Token Plan (Qwen (Alibaba), 11 Nov 2026)
- Alibaba opens Qwen4-Max preview with a million-token context window (TechNode, 11 Nov 2026)
- Qwen4-Max Preview: How to Try It, Token Plan Pricing and Specs (Neurohive, 11 Nov 2026)
- Qwen4-Max Preview: 1M Context, 15 Days, No Benchmarks (Gadget Pilipinas, 11 Nov 2026)
Reference pages
Spotted an error? See the corrections policy and tell the editor. Facts in this story come from the sources listed above; vendor claims are attributed.
Comments
Loading comments.





