Chinese e-commerce giant Alibaba has made its video and still image-creating AI model, Wan 2.1, publicly available, paving the way for its mass deployment and fueling competition in the AI field.
Image source: Alibaba
Open-sourcing AI models is a common move in the AI industry, with DeepSeek being one of the most prominent players. Alibaba has released four variants of Wan 2.1: T2V-1.3B, T2V-14B, I2V-14B-720P, and I2V-14B-480P, which generate videos and still images based on a text query or a template, which can be an image. The “1.3B” and “14B” designations indicate that these variants contain 1.3 billion and 14 billion parameters, respectively.
The models are available to users worldwide on the HuggingFace and ModelScope (part of Alibaba Cloud) platforms for academic, research, and commercial use. Alibaba released its latest version of its video-generating AI model in January, initially called Wanx but later renamed Wan. The project has scored highly in Vbench, a benchmark for video generators, including a top score for object interactions.
Alibaba also released a draft of the QwQ-Max reasoning model the day before, which will also be released as an open-source project later. The company plans to invest at least 380 billion yuan ($52 billion) in the next three years to support cloud computing and AI infrastructure.