Wan 2.1 T2V 14B

Wan 2.1 T2V 14B is Alibaba's open text-to-video model, released on February 25, 2025 under Apache 2.0. It is a 14-billion-parameter diffusion transformer with a novel 3D causal VAE for spatiotemporal compression, generating up to 81 frames at 480p or 720p from text prompts in Chinese and English.

At release it topped the VBench leaderboard at 86.22 percent in Alibaba's evaluation, ahead of commercial systems of the time, and it was the first major open video model with native Chinese text-to-video. The release included 14B and 1.3B sizes plus image-to-video variants, with diffusers and ComfyUI integrations landing shortly after.