Fact · Languages and accents
78percent of 1,570 evaluated languages
Meta's Omnilingual ASR paper reports that its 7B model meets the character error rate threshold for 78% of the 1,570 languages it was evaluated on.
Checked against its source on October 17, 2026
In the source's own words
“overall our model meets the CER quality threshold for 78% of the 1570 languages we evaluate on”
Omnilingual ASR: Open-Source Multilingual Speech Recognition for 1600+ Languages, Omnilingual ASR team, arXiv, 2025, Section 5.3.
Open the original- What it measures
- The threshold is a character error rate below 10 percent. The paper evaluates on 1,570 languages, which is fewer than the 1,600 languages the model is said to cover.
- What it does not show
- The threshold is a coarse quality bar, and the authors' figures come from their own test sets. Fewer than 1,600 languages were evaluated.