Wan 3.0: What Is Officially Known and What Is Still Unconfirmed
Is Wan 3.0 officially released? We checked Wan's official repositories, Hugging Face organization, and Alibaba Cloud API docs to separate confirmed facts from third-party claims.
TL;DR
Wan 3.0 has not been officially confirmed as a public Alibaba or Wan model as of August 6, 2026. The official open-source line still points to Wan2.2, while Alibaba Cloud documentation uses Wan2.7 for its current video API.
- ✅ Use Wan2.2 if: you need an official open-source Wan repository, weights, and license information
- ✅ Use Wan2.7 if: you need a documented Alibaba Cloud video API
- ❌ Treat Wan 3.0 claims carefully if: the page has no official model card, repository, weights, or API model ID
Is Wan 3.0 officially released?
As of August 6, 2026, we could not find an official Wan release named Wan 3.0 in the Wan team's public GitHub organization, the Wan-AI Hugging Face organization, or Alibaba Cloud's current Model Studio video API documentation. That means there is no official release date, model card, public weight download, license, or API identifier that we can verify for a Wan 3.0 model today.
That answer may feel less exciting than a new model announcement, but it is useful when you are deciding what to download, what to put into a production workflow, or what to pay for. A third-party page can use the name “Wan 3.0” without being an Alibaba or Wan release.
What is the latest official Wan model?
The answer depends on which part of the Wan ecosystem you mean.
| Area | Latest source we could verify | What it tells us |
|---|---|---|
| Open source | Wan2.2 official repository | Code, model weights, and a documented Apache 2.0 license |
| Alibaba Cloud API | Wan2.7 documentation | A cloud API with a wan2.7-t2v model name |
| Wan 3.0 | No official source found | Release status and specifications remain unconfirmed |
The Wan2.2 repository describes a mixture-of-experts design for the two diffusion stages and includes text-to-video, image-to-video, and text-image-to-video releases. Its repository records the code and weights release on July 28, 2025, under the Apache 2.0 license. Those are concrete details a developer can inspect and use.
Alibaba Cloud's Model Studio documentation uses the wan2.7-t2v name for text-to-video. The current reference describes text prompts, 720p and 1080p output options, and durations from 2 to 15 seconds. This is a cloud service version. It should not be treated as proof that a Wan3.0 open-source model already exists.
Why are people talking about Wan 3.0?
Model names travel quickly once creators, API aggregators, and video tools start discussing a possible next version. A product page may use a forward-looking label, a provider may rename an endpoint for its own catalog, or a search result may merge several Wan-related products under one phrase.
Those pages can still be useful for discovering what people are trying to use. They do not settle the question of official release status. For that, the name needs to connect to an official repository, model card, weight file, license, or API document owned by Alibaba or the Wan team.
This is especially important for a video model. A name can imply capabilities that the underlying service does not provide. Resolution, duration, image inputs, reference controls, audio, pricing, and commercial rights all need their own source. One label cannot verify them all.
How to verify a Wan model page
Before you download a checkpoint or build a workflow around a new model name, check five things.
- Repository. Is there an official repository under the Wan team's GitHub organization?
- Model card. Does the official card describe inputs, outputs, limitations, and intended use?
- Weights. Is there an official download location with a clear version name?
- License. Can you see the license and its commercial-use terms?
- API identifier. If the page is an API product, does the provider publish the exact model ID in its own documentation?
If a page has none of these and only repeats a model name, keep it in the “unconfirmed” column. That is a better working habit than copying a specification table that nobody can trace back to its owner.
Which Wan version should you use today?
Choose the route that matches the work in front of you.
| Your situation | Practical starting point |
|---|---|
| You want to inspect code and run an open model | Wan2.2 official repository |
| You want a documented hosted API | Alibaba Cloud's Wan2.7 documentation |
| You found a third-party Wan 3.0 page | Check its provider, exact API ID, limits, and source before relying on it |
| You are planning future content | Write about the Wan family and update the article when Wan3.0 is officially documented |
The safest article title today is also the most useful one. Explain what has been confirmed, show readers where to check, and leave room to update the page when a formal Wan3.0 release appears.
Try a Wan video workflow
You can explore the current Wan video workflow in the Wan video generator. The page is useful for testing prompts and planning shots, while the official Wan2.2 and Wan2.7 sources remain the right references for release status and model documentation.
Related Reading
- Wan AI Video Prompt Guide — Turn a broad idea into a shot the model can interpret
- Wan AI Video Workflow — Move from a short story idea to a usable shot list
- Wan2.7 video generator — Explore the current Wan video experience
FAQ
Disclosure
Release-status information was checked against the Wan-Video GitHub organization, the Wan-AI Hugging Face organization, the Wan2.2 repository, and Alibaba Cloud's Wan2.7 text-to-video API reference on August 6, 2026. Product names and API availability can change.
더 많은 기사
Wan AI Video Prompt Guide: Turn a Simple Idea Into a Usable Shot
Learn how to write Wan AI video prompts with clear subjects, actions, camera direction, lighting, and constraints. Includes reusable text-to-video and image-to-video templates.
Wan AI Video Workflow: From One Story Idea to a Coherent Short
A practical Wan AI video workflow for turning a story idea into characters, shots, prompts, generated clips, and a finished edit without losing continuity.