Cohere launches Transcribe, its first voice model; the 2B-parameter, open-source speech recognition model handles tasks like notetaking and speech analysis TechCrunch 2026-03-27 Ivan Mehta Cohere Related Coverage Introducing Cohere Transcribe: a new state-of-the-art in open-source speech recognition Cohere Excited to share that our team at Cohere is launching a new state-of-the-art ASR model. — It ranks #1 for English accuracy on Hugging Face's Open ASR Leaderboard … Shuangyi Cohere's Open-Source Transcribe Model Tops ASR Leaderboard WinBuzzer · Markus Kasanmascheff