← Glossary Term
FLEURS
A multilingual speech benchmark used to test how accurately a model transcribes audio across many different languages.
FLEURS is a standard test set for speech recognition, built from the same sentences recorded in a long list of languages. Because the content is held constant and only the language changes, it shows whether a transcription model is genuinely multilingual or just very good at English.
It matters when you read accuracy claims. A speech model might report a word error rate of two or three percent, which sounds excellent, and then score noticeably worse on FLEURS once other languages are included. If you do not work in English, the FLEURS number is the one to look at.
Mentioned in