MMS-1B
Meta's 1B speech model covering 1,100+ languages for ASR and TTS.
About
MMS-1B is the 1-billion-parameter flagship of Meta's Massively Multilingual Speech (MMS) project, announced on May 22, 2023. Built on wav2vec 2.0 self-supervised pretraining over 500,000 hours of speech in 1,400 languages, it provides speech-to-text and text-to-speech in 1,107 languages plus language identification across 4,000 spoken languages.
Meta reported half the word error rate of OpenAI's Whisper with 11x the language coverage, using religious texts like the Bible as parallel data for low-resource languages. The models and code were released openly for research, positioned as infrastructure for language preservation rather than a commercial product. License verify-card: confirm the non-commercial terms before production use.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.
Use cases
Trust score — 8/10
Scored 8/10 (verified tier, metrics-v1). Evidence: Verified via vendor site/docs/GitHub and reputable coverage, wave 112; Trending score 84/100 from researcher assessment.