Alibaba Qwen ships Qwen-Audio 3.1 voice model with up to 50% price cut across voice APIs
Alibaba’s Qwen team released Qwen-Audio 3.1 on September 27 with accuracy and low-SNR improvements, and cut API pricing across the full Qwen-Audio voice line by up to 50%.
NetEase / Qianyan Online aggregation on September 27 reports that Alibaba’s Qwen team released Qwen-Audio 3.1 on September 27 with improvements in recognition accuracy, language coverage, low-SNR performance, and speaker diarization. The release also cut API pricing across the full Qwen-Audio voice line by up to 50 percent.
On the technical side, Qwen-Audio 3.1 brings a systematic update to accuracy, low-SNR performance, and speaker separation, and extends support for multilingual, dialect, and code-switched audio. Alibaba notes that Qwen-Audio 3.1 shares the underlying training framework with the company’s upcoming multimodal models, allowing customers to share context and prompts across text, vision, and voice workloads.
On pricing, Qwen-Audio voice APIs are cut by up to 50 percent. The cut is consistent with the low-price strategy of Qwen3.8-Flash in international markets and is being read as a deliberate move to widen the affordability of voice AI following the September 24 Apsara Conference.