Alibaba has made its advanced AI model Wan 2.1 open source, allowing public access to its video- and image-generating capabilities. This move is expected to increase the model’s adoption and intensify competition in the AI sector. The company has released four variants of Wan 2.1, including T2V-1.3B, T2V-14B, I2V-14B-720P, and I2V-14B-480P, which can generate images and videos from text and image inputs. The “14B” variants can process up to 14 billion parameters, resulting in more accurate outputs.
Wan 2.1 is now available globally on Alibaba Cloud’s ModelScope and HuggingFace platforms for academic, research, and commercial use. This release follows Alibaba’s introduction of the latest version in January, which was praised for its ability to generate highly realistic visuals. The model has gained recognition for its top ranking on VBench, a leaderboard for video generative models, particularly excelling in multi-object interaction functionality.
In addition to Wan 2.1, Alibaba has also previewed its reasoning model QwQ-Max, which the company plans to make open source upon full release. This move by Alibaba follows a similar action by startup DeepSeek, whose open-source models have recently generated excitement among technology investors due to their performance comparable to more established competitors like OpenAI.
Answer from Perplexity: pplx.ai/share
