Intel / post

Hy ASR 3.0 preview claims ~3% multilingual word error rate

Source: x.com
Date: Published Aug 4, 2026

Context

Tencent's Hy ASR 3.0 preview converts speech into text. Its announcement claims a multilingual word error rate of roughly 3%, and a company-published chart gives 3.34% for Chinese, 2.62% for English and 3.12% for Cantonese. Lower values mean fewer transcription errors.

The chart identifies public speech datasets used for each language, but does not explain how their results were combined. These are reported results for those evaluations; they do not establish a single error rate for every language, speaker or recording environment.

Terms in this piece · Glossary

benchmark — A standard public test set for comparing AI models — the shared scoreboards behind every "model X beats model Y" claim.
LLM — A large language model — the neural network behind tools like Claude and ChatGPT, trained on huge amounts of text to predict what comes next.
context window — The maximum amount of text a model can consider at once — its working memory for the current conversation or task.

Read the source x.com