Wan 3.0 Prime

• 🏢 Alibaba's premium text-to-video tier in the third-generation Wan family.
• 🆕 Released August 2026 alongside image-to-video and reference-to-video Prime variants.
• 📏 Alibaba says Wan 3.0 generates up to 30 seconds in one pass.
• 🌐 Inputs can include text, images, audio, video and documents.
• 🔧 One all-in-one model spans text, image and reference-driven video tasks.
• 🎯 Catalog positions it for photorealism and subject coherence across frames.
• ⚡ Served through Alibaba Cloud Model Studio's video generation APIs.
• 📚 Follows the earlier Wan 2.7 and Wan 2.6 video generations.

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 premium text-to-video tier of Alibaba's third-generation Wan video family, listed here alongside Wan 3.0 Prime for image-to-video and Wan 3.0 Prime Reference. In this catalog it targets photorealistic clips where a subject, prop or setting must stay stable as a shot develops.

The structural change comes from the base generation, Wan 3.0. Alibaba describes Wan 3.0 as an all-in-one system that handles text-to-video, image-to-video and reference-driven generation inside a single model, and that extends single-pass output to about 30 seconds. Input coverage widens too: beyond text and stills, Alibaba says generation can be conditioned on audio, existing video and documents, so a brief or a deck can seed a clip.

Compared with Wan 2.7 and Wan 2.6, the differences are scope and interface. Wan 2.7's API reference had already reworked control surfaces, replacing a dedicated shot-type parameter with prompt-level shot description and swapping fixed pixel sizes for a resolution tier plus aspect ratio. Earlier 2.1 through 2.6 image-to-video work was documented as separate first-frame endpoints. Wan 3.0 consolidates those task-specific paths and lengthens the usable single take.

Prime sits above the standard Wan 3.0 entry as the higher-fidelity option for the same text-to-video task, accessed through Alibaba Cloud Model Studio's video generation service.