Cohere Releases Cohere Transcribe

Cohere, an enterprise artificial intelligence company, has entered the automatic speech recognition (ASR) market with its release Cohere Transcribe, an open-source ASR model for tasks like note-taking and speech analysis.

The lightweight Cohere Transcribe, which contains just 2 billion parameters, currently supports 14 languages: English, French, German, Italian, Spanish, Portuguese, Greek, Dutch, Polish, Chinese, Japanese, Korean, Vietnamese, and Arabic. The model reportedly achieved a 5.42 percent word error rate and can process up to 525 minutes of audio in a minute.

Cohere Transcribe also offers a robust chunking and reassembly logic that can process audio in 35-second segments. For files that exceed this limit, the system automatically splits the audio into overlapping chunks, processes each segment through its conformer-transformer pipeline, and then reassembles the overlapping text.

Cohere said it plans to integrate Transcribe into its North agent orchestration platform and is making the model available through its API and the Model Vault managed inference platform.