Cohere Transcribe: open-weights ASR topping the Open ASR leaderboard

Cohere Transcribe turns recorded speech into written text across 14 supported languages. Its documentation describes a dedicated speech-recognition model released under Apache 2.0. The company also describes local deployment, giving organizations a way to process audio on infrastructure they control.

Its scope matters when building a transcription workflow: the documented model expects a specified language and does not provide timestamps or identify individual speakers. It supplies the spoken words; applications needing a timed transcript or speaker labels must add those capabilities separately.