Search
On the morning of March 27, Alibaba introduced Tongyi Qianwen Qwen2.5-Omni-7B, its groundbreaking end-to-end omnimodal large model, and made the bold announcement that it would be open-sourced. This model is capable of concurrently processing a diverse array of inputs, including text, images, audio, and video, and delivering real-time text and natural speech synthesis outputs. In evaluations for multimodal fusion tasks, such as those conducted on OmniBench, Qwen2.5-Omni set new benchmarks, surpassing comparable models like Google's Gemini-1.5-Pro. Qwen2.5-Omni's ability to perceive the world mimics human multisensory perception, enabling seamless real-time interaction. Furthermore, it can discern emotions through audio and video, offering smarter and more intuitive feedback and decision-making in complex scenarios.
-
C114 Communication Network -
Communication Home
