Alibaba announced that its AI models, which can generate photos and videos from text prompts or images, are becoming free. The company is also sharing the open source code of some of the models with scientists, researchers, and commercial institutions around the world.
The four models are from the Wan2.1 family, the latest version of the Chinese company's core AI model. Their open source code will be available through Alibaba Cloud Model Scope and the Hugging Face AI model repository.
Users will be able to use T2V-14B, T2V-1.3B, I2V-14B-720P and I2V-14B-480P. 14B and 1.3B. The model names indicate the number of parameters they can perform, namely 14 and 1.3 billion, respectively.
The Wan2.1 family was introduced in early 2025, becoming the first AI models capable of generating video, supporting both Chinese and English input. They topped the VBench ranking, which evaluates AI for video generation, with an overall score of 86.22%. They also lead in parameters such as dynamic range, spatial relationship, color, and interaction between multiple objects.
We remind you that it was previously reported that Alibaba will collaborate with Apple on artificial intelligence for the iPhone in China. In addition, the company introduced a new artificial intelligence model, which, according to representatives, is superior to DeepSeek.