Alibaba has confirmed that Qwen 4 is in development. Its latest teaser gives readers a reason to watch for news, but it does not establish a release date, a downloadable model or an API you can use.
On November 6, Qwen’s official account posted “The becoming begins.” The accompanying image carries Qwen branding and the same English line, but neither the post’s text nor its image names Qwen 4. Reading it as a Qwen 4 launch signal is an interpretation, not a product announcement.
This page separates the confirmed record from the questions that still need answers. Status checked November 6, 2026: we have not verified a Qwen 4 launch date, public API model ID, price or official downloadable checkpoint in the sources listed below.
Qwen 4 release status at a glance
| Reader question | What the checked evidence establishes |
|---|---|
| Has Alibaba named Qwen 4? | Yes. Its September 22 announcement said Qwen 4 was in training. |
| Does the November 6 teaser confirm a launch date? | No. It gives no date and does not name the model. |
| Can I download Qwen 4? | We did not verify an official Qwen 4 checkpoint in our checks of Qwen’s Hugging Face listings. |
| Can I call a Qwen 4 API? | We did not verify a Qwen 4 model ID in the Model Studio release information checked for this article. |
| What does it cost? | No Qwen 4 price was verified in this review. Existing Qwen model prices do not establish its price. |
| Will it run locally? | Unresolved. A useful answer needs an exact released checkpoint, file sizes, license and supported runtime. |
The availability checks above are bounded observations of Qwen’s official Hugging Face organization and Alibaba Cloud Model Studio’s model lifecycle page. They do not rule out private testing or an announcement on a surface we could not inspect.
What Alibaba has confirmed
In its September 22, 2026 Apsara announcement, Alibaba described Qwen 4 as its next-generation model and said it was in training. That establishes the project’s existence at that date. It does not tell us when training finishes or when access opens.
The same announcement discussed Qwen 4.5 and Qwen 5 scaling to 5–10 trillion parameters. Those roadmap numbers belong to the later model series. They are not confirmed Qwen 4 specifications.
For the November teaser, the useful distinction is between an official source and an explicit statement. The account is official; the inference about which model it is teasing remains an inference. A countdown, named release or working access route would provide a firmer basis for a launch claim.
Qwen3.8-Flash-Next is a separate released model
There is already a downloadable checkpoint named Qwen/Qwen3.8-Flash-Next. Its Hugging Face page displays the qwen4_exp architecture tag. That tag can explain some confusion, but it does not rename the released checkpoint Qwen 4.
Keep the exact identity attached to any result you read. A benchmark, memory figure or license for Qwen3.8-Flash-Next describes that release. It cannot establish the same property for a future Qwen 4 checkpoint. Kingy’s Qwen3.8-Flash-Next hardware guide covers that existing model and preserves the distinction.
What to check when Qwen 4 arrives
The practical questions depend on how you plan to use it. The diagram below is a verification sequence, not a promised rollout order.
For app users, look for an announcement naming the model and saying which accounts, plans and regions receive it. A model appearing in a demonstration does not establish that your account has access.
For API developers, record the exact model ID, deployment region, context limits and input/output billing units. Test the documented route before changing a production dependency. A future price comparison should keep cached input, ordinary input and generated output separate wherever the provider bills them differently.
For local users, start with the official checkpoint and license. Then check weight-file sizes and a supported runtime. Parameter counts alone are insufficient for a buying decision: the intended context length, precision and runtime overhead also affect the memory needed. We cannot give Qwen 4 a hardware verdict before those details are available.
For performance claims, preserve the model version, benchmark version, reasoning settings and tool setup. Vendor-reported results should remain attributed to the vendor. Kingy will need a disclosed test setup before describing any result as its own measurement.
Limits and update policy
Not tested by Kingy. This is a source-based launch-status article. We have not run Qwen 4, measured its speed or independently reproduced performance claims. Qwen’s blog landing page returned no readable article text to our research tool, so it is not evidence for an exhaustive review of every Qwen announcement.
Updates to this page will identify the material fact that changed and link its source. A confirmed release date, working access route, official checkpoint, license, price or documented specification can justify an update. A source recheck with the same result does not justify a new updated date.
Sources
- Qwen’s official teaser, November 6, 2026; inspected directly in the browser.
- Alibaba’s full-stack AI roadmap announcement, September 22, 2026.
- Qwen’s official Hugging Face organization and Qwen3.8-Flash-Next model card.
- Alibaba Cloud Model Studio: model lifecycle and updates.
Sources checked November 6, 2026. Availability observations can change after that check.
Revision log
November 6, 2026: Initial publication, separating Alibaba’s confirmed development statement from the latest teaser and unverified release details.
Trending on Kingy
Keep reading with the stories getting the most attention now.