Chinese e-commerce giant Alibaba announced on Wednesday that its video- and image-generating AI model, Wan 2.1, is now open source. The move is expected to drive wider adoption and intensify competition in the AI sector.
Alibaba’s decision follows a similar move by startup DeepSeek, whose low-cost open-source models have drawn significant interest from technology investors. DeepSeek’s performance has surprised many in the capital-intensive AI space, matching the capabilities of established players like OpenAI.
Four Variants of Wan 2.1 Released
Alibaba has made four variants of Wan 2.1 publicly available:
- T2V-1.3B
- T2V-14B
- I2V-14B-720P
- I2V-14B-480P
These models generate images and videos from text and image inputs. The “14B” designation signifies that those variants process 14 billion parameters, allowing for more precise and detailed outputs.
The models are accessible worldwide through Alibaba Cloud’s ModelScope and the Hugging Face platform for academic, research, and commercial use.
Advancing AI Capabilities
Alibaba introduced the Wan 2.1 model in January, initially under the name Wanx, before shortening it. The company has emphasized its ability to produce highly realistic visuals, earning it a top ranking on VBench, a benchmark for video-generation models. Wan 2.1 leads in key areas such as multi-object interaction.
On Tuesday, Alibaba also previewed its reasoning model, QwQ-Max, which it plans to make open source upon full release.
Major Investment in AI and Cloud Computing
In addition to its AI developments, Alibaba announced plans this week to invest at least 380 billion yuan ($52 billion) over the next three years to enhance its cloud computing and AI infrastructure.
This significant investment underscores Alibaba’s commitment to advancing AI technology and maintaining a competitive edge in the global market.
