Cohere deployed Transcribe, an open-source automatic speech recognition model that converts spoken audio into text across 14 languages. It is available as a free API, a downloadable open-source model, and as a managed enterprise service through Cohere's Model Vault platform.

Cohere Transcribe (released March 26, 2026) is a 2-billion-parameter encoder-decoder model that takes audio input and outputs text transcripts. It supports 14 languages including English, French, German, Chinese, Japanese, Korean, and Arabic, and claims the lowest word error rate on the Hugging Face Open ASR Leaderboard at 5.42% as of its launch date. Cohere plans to integrate Transcribe into its North enterprise agent platform to add voice capabilities to automated workflows. The model is available for download on Hugging Face under an open-source license, via the Cohere API for free up to rate limits, and through Model Vault for production deployments.