Alibaba Open-Sources AI Model Wan 2.1 to Boost Competition in Video & Image Generation

Alibaba (NYSE:BABA) has officially open-sourced its video- and image-generating AI model, Wan 2.1, intensifying competition in the AI landscape. This move follows a similar release from DeepSeek, a startup that gained attention with its cost-effective models rivaling industry giants like OpenAI.

Wan 2.1 comes in four variants—T2V-1.3B, T2V-14B, I2V-14B-720P, and I2V-14B-480P—designed to generate images and videos from text and image inputs. The "14B" models leverage 14 billion parameters, enhancing accuracy and realism. These models are now accessible worldwide on Alibaba Cloud’s ModelScope and Hugging Face, catering to academic, research, and commercial users.

Originally launched in January under the name Wanx, the model was later rebranded as Wan. It has since secured the top spot on VBench, a leading benchmark for video generative models, excelling in multi-object interaction and complex scene rendering.

Alibaba is further expanding its AI capabilities with QwQ-Max, an advanced reasoning model previewed this week, set for open-source release soon. To strengthen its AI and cloud infrastructure, the company has pledged a massive 380 billion yuan ($52 billion) investment over the next three years.

By making Wan 2.1 open source, Alibaba is positioning itself as a major player in the AI race, challenging rivals and accelerating innovation in video and image generation technology.