Qwen Unveils Wan2.2, a 14B Open Text-to-Video Model

The Qwen team at Alibaba has introduced a significant new open-source model for generating video from text prompts, called Wan2.2-T2V-A14B. This release expands the team's portfolio of powerful, openly accessible foundation models.

What sets this model apart is its use of a Mixture-of-Experts (MoE) architecture. It features 14 billion active parameters, meaning only a fraction of the model's total size is engaged for any given task. This design aims to deliver the performance of a much larger model while keeping the computational cost of inference more manageable.

The release of Wan2.2 is a notable event in the competitive landscape of generative video. By making the model available under the permissive Apache 2.0 license, the Qwen team provides researchers and developers with a powerful, unrestricted tool for building new applications and pushing the boundaries of video synthesis.

The model is now available for download and experimentation on the Hugging Face Hub, allowing the community to begin exploring its capabilities immediately.