Speech interaction is experiencing a new technological breakthrough. Recently, the Qwen large model officially announced an upgrade to its core audio capabilities, launching a new real-time speech recognition model called Fun-ASR-Realtime.

The model has achieved significant breakthroughs in performance metrics, with the first character recognition delay accurately controlled within the hundred-millisecond range, ensuring users experience a smooth "speak and get feedback immediately" interaction. In terms of accuracy, the model performs excellently, with its recognition level approaching that of industry-leading offline recognition models, truly achieving an organic integration of real-time performance and high quality.

To meet diverse application needs, Fun-ASR-Realtime offers extensive language coverage. It not only supports the recognition and processing of up to 30 languages, but also particularly enhances the understanding of Chinese dialects, currently capable of recognizing complex voice inputs including 16 dialects.

This technological upgrade not only provides a more solid underlying support for real-time interaction scenarios such as voice assistants and meeting notes, but also further expands the application boundaries of general large models in multilingual and multi-dialect environments. With the official launch of the model, developers and enterprise users will be able to more conveniently build intelligent voice applications with high sensitivity and accuracy, making AI interactions more natural and human-like.