Woofun AI reports that Alibaba has released Qwen-Audio-3.0-Realtime, a speech model designed for millisecond-level responses and active invocation of external tools like maps, APIs, MCP, and knowledge bases. The system operates without explicit user commands for tool usage, retaining context across queries to refine search results, such as filtering restaurant recommendations by rating and distance. It supports simultaneous listening and speaking, allowing users to interrupt at any time, while employing speaker identification to mitigate background noise in multi-party or noisy settings. The model dynamically adjusts tone, pace, and emotion based on conversational context. Both the Plus version for enhanced inference and the Flash version for faster response times are now available on Alibaba Cloud Hundred Refinements.