Alibaba launches Wan3.0 video model with 30-second clips

Alibaba's Tongyi Lab officially released its Wan3.0 video model on 24 August 2026, following a public beta that began on 6 August. Alibaba said the model can generate clips up to 30 seconds long, double the 15-second limit of its predecessor, Wan2.7-Video.

The model produces video at 480p, 720p or 1080p resolution, with audio generated in the same pass by default, according to Alibaba. It renders realistic faces with synchronised micro-expressions and voices in several languages, and an "intelligent duration" feature suggests a clip length, with an option to extend it.

Wan3.0 can also convert documents of up to 50 pages, spreadsheets, slides and web pages into video, Alibaba said. A feature called "Omni-Reference" takes up to 20 references, including images, video, audio, documents and web pages, and keeps them consistent across a generated clip.

Alibaba priced API access at $0.05 per second at 480p, $0.10 at 720p and $0.20 at 1080p, or $12 a minute. That compares with $0.40 per second for the standard tier of Google's Veo 3.1.

Wan3.0 is a closed model available only through Alibaba's Model Studio API, which is in preview. Alibaba's open-weight Wan line stops at Wan 2.2.

The release came days after Alibaba raised $10 billion in a share placement to fund AI spending, a deal Reuters called the largest-ever primary follow-on offering by a Hong Kong-listed company, according to TechNode.

Alibaba cited uses including short dramas and film, advertising, tourism, marketing, music videos, and simulated video for training self-driving and robotics systems, according to The Next Web.