Alibaba Qwen Releases Qwen-Audio-3.1-Realtime for Voice Agents

Qwen launched the Qwen-Audio-3.1 stack of five models spanning ASR, TTS and realtime interaction. The flagship Qwen-Audio-3.1-Realtime is a full-duplex speech model with function calling and web search, a 262K-token context and $6.4 per 1M audio input tokens. No open weights; access is via the QwenCloud API.

• 262K-token context: 245K input, 16K output
• $6.4 per 1M audio input and $24 per 1M output tokens
• Price cuts: ~85% on Realtime, up to 95% on ASR
• FLEURS WER dropped from 9.01 to 3.98