Tongyi Lab has launched its latest speech recognition model, Fun-ASR 1.5, on April 20. The model, now available via Alibaba Cloud's Bailian and the ModelScope community, supports 30 languages, seven major Chinese dialect groups, and over 20 regional accents with a single model. This eliminates the need for separate models for each dialect. Internal tests show a 56.2% reduction in character error rate for dialect scenarios compared to the previous version, with five dialects achieving over 90% accuracy. The model also features enhanced recognition for classical poetry, boasting a 97% character-level accuracy. This unified system addresses the long-tail challenge of Chinese dialect speech recognition, making it commercially viable for applications like educational live streaming, local government hotlines, and interview transcription, simplifying deployment by removing the need for multiple recognition pipelines.