Translated data: Alibaba's research team has recently launched the Qwen-Audio series, successfully addressing a variety of challenges in the field of large-scale audio language models. This series covers a wide range of audio types, including human speech, natural sounds, music, and songs, and has achieved outstanding performance across multiple tasks without the need for task-specific fine-tuning. Qwen-Audio-Chat, as an extension, demonstrates universal audio comprehension capabilities, supporting multi-turn conversations and various audio-centric scenarios. Alibaba states that this breakthrough fills the gap in large-scale language models' understanding of audio, expanding pre-training to 30 tasks and a variety of audio types.