Introduction
Alibaba has launched Wan3.0, a model capable of generating videos from text, images, audio, video, PDF, presentations, and web pages. The model can generate content up to 30 seconds long in a single run and at 1080p resolution. Doubling the 15-second limit of the previous version, Wan2.7, Wan3.0 can also generate multilingual audio outputs and movements synchronized with facial expressions.
Developments
Available through Model Studio and Qwen Cloud, the model is priced per second at $0.05 for 480p, $0.10 for 720p, and $0.20 for 1080p. Unlike previous Wan models, Wan3.0’s weights will remain closed, while API access is planned to be rolled out gradually.
The launch followed Alibaba’s approximately $10.2 billion share sale in Hong Kong.
Details
The company said it would allocate the proceeds to chips, artificial intelligence infrastructure, and model development. In early 2025, Alibaba announced plans to invest 380 billion yuan in artificial intelligence infrastructure over three years. The company’s cloud and artificial intelligence revenue rose 45% to 48.44 billion yuan in the June quarter.