Alibaba Open-Sources AI Model for Video and Image Generation, Intensifying Competition in the AI Industry

Alibaba has released its video- and image-generating AI model, Wan 2.1, open source, which is intended to boost adoption and increase competition in the AI field.  The statement follows similar efforts taken by company DeepSeek, whose low-cost models have sparked investor interest by outperforming established AI leaders such as OpenAI.  

The four variants of Wan 2.1—T2V-1.3B, T2V-14B, I2V-14B-720P, and I2V-14B-480P—can produce images and videos from text and image inputs.  The "14B" variants process 14 billion parameters, resulting in more sophisticated and precise visual outputs.  The models are now available on Alibaba Cloud's ModelScope and Hugging Face for academic, research, and commercial purposes.  

Alibaba first released Wan 2.1 in January, shortening its name from Wanx, and has since emphasized its high-quality images.  The model achieved the top ranking on VBench by excelling at multi-object interaction for video creation.  

Beyond AI models, Alibaba is growing its AI and cloud infrastructure by committing 380 billion yuan ($52 billion) over the next three years.  The company also previewed their new reasoning model QwQ-Max, which would be open-source once fully released.  

With these developments, Alibaba is establishing itself as a prominent player in the global AI competition.