Search
According to information from Tongyi Large Model, on December 23, 2025, Alibaba made a significant move by releasing its latest end-to-end speech interaction model, Fun-Audio-Chat. In a generous gesture towards the tech community, Alibaba has open-sourced not only the 8B model weights but also the inference code and Function Call integration examples. This cutting-edge model has emerged as a frontrunner, claiming the top spot among models of similar size across multiple leaderboards. Its overall performance is outstanding, leaving behind competitors such as GLM4-Voice and Kimi-Audio. What sets Fun-Audio-Chat apart is its adoption of an end-to-end S2S (Speech-to-Speech) architecture. This design choice translates into enhanced efficiency and significantly reduced latency, making it a game-changer in the field of speech interaction.
-
C114 Communication Network -
Communication Home
