NNewsGPT ← Home
CN

Tencent Hunyuan Unveils New Hy ASR 3.0 Speech Recognition Model

CN1 hr ago

Tencent Hunyuan has officially launched its latest generation speech recognition model, Hy ASR 3.0 preview. The model demonstrates strong performance across multiple languages, achieving a Word Error Rate (WER) of approximately 3% on various open-source evaluation datasets. Specifically, Hy ASR 3.0 preview recorded a WER of 3.34% for Mandarin Chinese, 2.62% for English, and 3.12% for Cantonese. This release marks a significant advancement in Tencent's speech recognition technology, aiming to improve accuracy and usability in diverse linguistic contexts.

AI Analysis

The release of Tencent Hunyuan's Hy ASR 3.0 preview highlights the rapid advancements in large-scale speech recognition models. Achieving low WER across multiple languages, including Mandarin, English, and Cantonese, suggests sophisticated acoustic and language modeling techniques. Such progress is crucial for enhancing human-computer interaction and enabling more seamless integration of voice interfaces across global markets. The competitive landscape for AI-driven speech technology continues to intensify, pushing for greater accuracy, lower latency, and broader language support, which will likely shape future communication platforms and accessibility tools.

AI-generated to prompt reflection — not editorial opinion, not advice, not a statement of fact. How this works.

Compiled by NewsGPT from 36Kr (CN). Read the original for full details.
ⓘ AdTurn your crypto wallet into a credit cardTurn crypto wallet → credit card · 50% spendable credits