Realtime AI News
Qwen Releases Qwen3-ASR-1.7B Automatic Speech Recognition Model
The Qwen team has officially released the Qwen3-ASR-1.7B model on Hugging Face, focusing on automatic speech recognition with support for Chinese, English, and Cantonese.
On June 26, the Qwen team published the Qwen3-ASR-1.7B model on the Hugging Face model registry, an open-source model dedicated to automatic speech recognition (ASR). Built on the Transformers architecture with safetensors format, it supports both text generation and speech recognition tasks.
According to the model tags, Qwen3-ASR-1.7B supports three languages — Chinese (zh), English (en), and Cantonese (yue) — covering a broad range of application scenarios. The model has received 4 likes since publication, representing a significant expansion of the Qwen3 series into the speech domain.
This release signals Qwen's continued investment in speech recognition technology. Qwen3-ASR provides developers and enterprises with a new open-source option for Chinese and Cantonese speech recognition, potentially accelerating the deployment of voice-interactive applications.
Why it matters
The open-source release of Qwen3-ASR provides a new foundational model for Chinese and Cantonese speech recognition, lowering the barrier for voice application development.