Alibaba releases next-generation end-to-end voice interaction model: Fun-Audio-Chat

Alibaba Tongyi has released its new-generation end-to-end voice interaction model, Fun-Audio-Chat. The model employs an innovative end-to-end sequence-to-sequence architecture, enabling direct generation of voice output from voice input, eliminating the need for traditional ASR+LLM+TTS multi-module concatenation and significantly reducing latency. In multiple authoritative benchmarks such as OpenAudioBench and MMAU, the model ranks first among models of the same size, with overall performance surpassing mainstream products such as GLM4-Voice and Kimi-Audio.