Wan 3.0 Prime

• 🆕 Prime tier of Alibaba's Wan 3.0 video generation, listed August 2026
• 👁️ Animates a supplied still image into a clip from a prompt
• 📏 Wan 3.0 generates up to 30 seconds in a single pass
• 🔧 Family covers first-frame and first-and-last-frame image-to-video tasks
• 🌐 Wan 3.0 is positioned as generating from a wide range of inputs
• 🎯 Emphasis on subject and scene coherence across frames
• 🏢 Served through Alibaba Cloud Model Studio
• 💬 Siblings cover text-to-video and reference-to-video variants

Alibaba

Alibaba Group is a Chinese multinational technology company founded in 1999 and headquartered in Hangzhou, Zhejiang. Originally built around e-commerce and cloud computing, Alibaba has become one of the most prolific contributors to open-weight AI research,…

Wan 3.0 Prime is the image-to-video member of Alibaba's Wan visual-generation family. You supply a still image plus a text prompt, and the model animates it into a clip. Alibaba describes the Wan 3.0 generation as an all-in-one system spanning text-to-video, image-to-video and reference-based generation, with output up to 30 seconds produced in a single pass. On this catalog it sits alongside Wan 3.0 Prime for prompt-only generation and Wan 3.0 Prime Reference for identity-anchored work.

Against its own lineage, the step is mostly about duration and input breadth. The 2.1 through 2.6 image-to-video endpoints, represented here by Wan 2.6 and Wan 2.1 Pro, took a first-frame image and a text prompt and returned short clips. Wan 2.7 broadened that interface beyond a single starting frame, adding first-and-last-frame and video-continuation tasks. Wan 3.0 extends single-generation length to 30 seconds and is presented by Alibaba as accepting a wider spread of input types.

Alibaba also highlights consistency of characters, props and scenes across a long take as a focus of the Wan 3.0 release. Practically, the Prime tier targets shots that must hold a subject stable for longer than a few seconds, rather than short cuts stitched together afterwards.

Models in this line are served through Alibaba Cloud Model Studio, whose documentation is the source for the capability details above. Independent third-party evaluations of this specific tier were not available at the time of writing.