Alibaba Releases Qwen-Audio-3.0-Realtime, AI Voice Model That Actively Calls Tools

According to Beating, Alibaba released Qwen-Audio-3.0-Realtime today, a real-time voice AI model capable of millisecond-level response and proactively invoking external tools such as maps, APIs, MCP, and knowledge bases during conversations without explicit user command.

The model supports simultaneous listening and speaking, allowing users to interrupt at any time. In noisy environments or multi-speaker scenarios, it identifies the primary speaker to reduce false triggers from background noise. Alibaba offers two versions: a more powerful Plus version and a faster Flash version, both now available on Alibaba Cloud Bailian.