Published on November 10, 2025, this briefing describes Omnilingual ASR as an open-source set of speech-recognition models and a dataset for more than 1,600 languages, including 500 described as previously unsupported.
On November 10, 2025, the briefing reported on Meta’s Omnilingual ASR: an open-source set of automatic speech-recognition models and a dataset with stated coverage of more than 1,600 languages. It highlights that 500 were described as previously unsupported and presents the system as aimed at broad multilingual coverage.
To check the announcement and assess these claims, consult the original publication and official project materials; verify how language coverage was measured and what conditions apply to the data and models. If using AI to study or apply the material, avoid entering identifiable recordings or other personal data without authorization, and follow your organization’s privacy rules.