Back to Featured
IT Home · Artificial Intelligence2026-07-06 14:39July 6

Alibaba Qwen Upgrades Fun-ASR-Realtime Speech Recognition Model, Supporting 16 Dialects and 30 Languages in a Single Model

One Model Supports 30 Languages and 16 Dialects: Alibaba Qwen Upgrades Fun-ASR-Realtime Speech Recognition Model

AI Briefing

Alibaba Qwen upgraded its real-time speech recognition model Fun-ASR-Realtime today. A single model supports 16 dialects and 30 languages, and its API is now available on Alibaba Cloud Bailian. In recognition tests covering 16 dialects across eight major dialect regions, it achieved an average character accuracy of 88.62%, leading related products from Volcano Engine and Tencent in 12 dialect categories. First-token latency is kept to the hundreds-of-milliseconds range, streaming recognition accuracy is close to offline performance, and the model supports seamless language switching. After targeted optimization for languages including Thai and other East and Southeast Asian languages, recognition accuracy improved by 20%. The offline model Fun-ASR-Flash has a 1.7% word error rate, ranking first globally on the AI evaluation platform Artificial Analysis.

Also reported by 1 other source

Source: IT Home · Artificial Intelligence · ithome.com

Finished reading the news? Let’s talk about your business.

Leave your contact details, and I’ll get in touch with you personally.

WECHAT

Add me on WeChat

Press and hold or select the WeChat ID below to copy it.