Wan3.0
Alibaba Cloud · China · 2026
Alibaba's video model that generates thirty seconds with sound in one pass and builds a clip straight from a Word file, a slide deck, a PDF or a web link - and the first Wan flagship with no published weights.
Wan3.0 is Alibaba Cloud's video model of August 2026 and the point at which the company's video line stopped being something a reader can download. It entered public beta on 6 August 2026 and launched on 24 August, a day after Alibaba raised roughly ten billion dollars in a Hong Kong share placement. The headline number is duration: thirty seconds generated in a single pass, twice the fifteen-second ceiling of Wan2.7, at 480P, 720P or 1080P, with sound produced together with the picture rather than laid over it afterwards. The more distinctive input is the one no rival takes. Alongside text, images, audio and video, Wan3.0 accepts documents - Word files, spreadsheets, slide decks, PDFs and Markdown - and public web links, reads their content and builds a video out of it. In the API this arrives as a media item of type file or type link, and the documentation is explicit that the two cannot be combined in one request. Alibaba also removed the shot-type switch used by earlier versions: the model now reads the prompt itself and decides whether the result should be a single take or cut between shots. Aspect ratio and duration can likewise be left adaptive. What the launch material does not say is the part worth knowing. Alibaba built the reputation of the Wan line on open weights - Wan2.1 in February 2025 and Wan2.2 in July 2025 were published under Apache 2.0 and remain among the most downloaded video models anywhere. Since then the served line has advanced through 2.5, 2.6, 2.7 and now 3.0, and not one of those four generations has weights on Hugging Face; the newest files on the company's account there are open releases built on the Wan2.2 base - Wan-Dancer in July 2026 and the Animate-2 character-animation weights in July and August 2026. The flagship is closed; the specialised work around it is not. The distance between what Alibaba runs and what Alibaba releases is now four generations wide, and Wan3.0 is the generation that made it obvious. A note on pricing: Alibaba documents that video is billed per output second and that the resolution tier changes the rate, but the company's price table is rendered by script and could not be read at source. The figures below come from trade reporting of the launch and should be treated as indicative.
▸News
▸Videos
No videos yet.
▸Reviews
No reviews yet. Be the first!