Tencent's new Hy ASR 3.0 speech model plugs into its Hy3 language model and hits about 3% word error rate across Mandarin, English, and Cantonese.
Continue to AI University →