Alibaba unveils Qwen-Audio-3.1 series and lowers prices across all products

On September 23, 2026, Alibaba's Qwen team launched the Qwen-Audio-3.1 series, a comprehensive audio model suite covering understanding, generation, interaction, and creation. The release includes five new models, with prices reduced across the board: TTS models saw a 70% price drop, Realtime models fell by 85%, and ASR models dropped by 95%. The Qwen-Audio-3.1-ASR model excels in multilingual and dialect recognition, achieving a 4.55% average CER and 10.38% for 16 dialects. It also supports structured transcriptions and semantic accuracy of 82.10%. The Qwen-Audio-3.1-ASR-Next model understands emotions, environmental sounds, and can answer audio-based questions. The TTS-Next model enables full audio creation, including voice, sound effects, and background noise, with support for 30 languages and 16 dialects. The Realtime model offers advanced conversational capabilities, including emotion recognition and agent integration, enhancing user interaction with AI.