Alibaba's AI model for video, image generation now publicly available
What's the story
Chinese e-commerce giant Alibaba has announced that its artificial intelligence (AI) model Wan 2.1 is now open-source.
The move is likely to boost its adoption and fuel competition in the AI sector.
The company has launched four versions of Wan 2.1 - T2V-1.3B, T2V-14B, I2V-14B-720P, and I2V-14B-480P - all of which are designed to create images and videos from text and image input.
Model features
Understanding the capabilities of Alibaba's AI model
The "14B" in the model names indicates that the versions can process up to 14 billion parameters, enabling them to handle more complex inputs and deliver better accuracy.
They are now available globally on Alibaba Cloud's ModelScope and HuggingFace platforms for academic, research, and commercial use.
The latest version of this video- and image-generating AI model was unveiled by Alibaba in January with an emphasis on creating hyper-realistic visuals.
Model performance
Leading in VBench rankings
Alibaba's AI model, formerly called Wanx, has been acknowledged for its best performance on VBench. The platform ranks video generative models according to their capabilities.
The company's model has taken a top spot owing to its advanced features such as multi-object interaction.
The achievement highlights the model's ability to generate highly realistic visuals, something that has contributed to its widespread adoption and use across fields.
Future developments
Alibaba previews new reasoning model and investment plans
Along with making Wan 2.1 open-source, Alibaba has also previewed a new reasoning model, QwQ-Max. The company plans to open-source the model when it fully launches it.
Alibaba also unveiled an ambitious investment plan for the next three years.
The company plans to invest at least CNY 380 billion ($52 billion) to bolster its cloud computing and AI infrastructure, showing its commitment to advancing in these fields.