Overview

Cohere Transcribe is a dedicated audio-in, text-out ASR model with a large Conformer encoder and a lightweight decoder. Cohere said at launch that it topped the Hugging Face Open ASR Leaderboard with an average word error rate of 5.42%.

Highlights

  • 2B parameters
  • 14 languages
  • Conformer-based encoder-decoder
  • Apache 2.0 license

Typical uses

Meeting and call transcription, subtitling and voice-data pipelines.

Things to consider

  • Leaderboard rankings change quickly; check current results.
  • Audio is resampled to 16 kHz mono, so stereo channels are mixed.

Specifications are taken from the model card on Hugging Face and the provider’s announcement. Benchmark results depend on evaluation settings; see how AI benchmarks work.