Alibaba (09988) topped the global open source model list with a thousand questions and answers.

On April 2nd, the world's largest AI open source community, Hugging Face, updated its list of large models. Alibaba's Qianyi Qianwen, recently open sourced by Alibaba (09988), reached the top of the overall list of large models with its end-to-end multimodal large model Qwen2.5-Omni, followed closely by DeepSeek-V3-0324 and Qunhe's SpatialLM-Llama-1B. This is the first time that Chinese technology companies have taken the top three spots on the global open source model list, highlighting Hangzhou's position as an AI innovation hub. The end-to-end multimodal large model Qwen2.5-Omni, which topped the list this time, can simultaneously handle various inputs such as text, images, audio, and video, and generate real-time text and natural speech synthesis outputs. Compared to closed-source large models with billions of parameters, Qwen2.5-Omni's small size of 7B makes it possible for the multimodal large model to be widely used in industry. It can even be easily deployed and applied on a smartphone. SpatialLM is a spatial understanding model developed independently by Qunhe Technology, which can generate physically correct 3D scene layouts based on just a video segment. Unlike traditional large language models, SpatialLM breaks through the limitations of understanding geometric and spatial relationships in the physical world, and will play a significant role in the spatial cognition and parsing capabilities of machine-like humans. In addition, the V3-0324 released by DeepSeek is a minor version update of V3. Although the official description is just a "minor version upgrade," its testing capabilities are close to the V3.5 version, especially in complex logic and multimodal understanding.