Introducing Cohere-transcribe: state-of-the-art speech recognition

Open paper
Benchmark Model Rank Results
speech-recognition-on-amiCohereLabs/cohere-transcribe-03-2026–Word Error Rate (WER): 7.8
speech-recognition-on-ami-cleanedCohereLabs/cohere-transcribe-03-2026–Word Error Rate (WER): 7.01
speech-recognition-on-earnings-22CohereLabs/cohere-transcribe-03-2026–Word Error Rate (WER): 10.38
speech-recognition-on-gigaspeech-cleanedCohereLabs/cohere-transcribe-03-2026–Word Error Rate (WER): 7.88
speech-recognition-on-gigaspeech-testCohereLabs/cohere-transcribe-03-2026–Word Error Rate (WER): 7.93
speech-recognition-on-librispeech-test-cleanCohereLabs/cohere-transcribe-03-2026–Word Error Rate (WER): 0.96
speech-recognition-on-librispeech-test-otherCohereLabs/cohere-transcribe-03-2026–Word Error Rate (WER): 2.04
speech-recognition-on-spgispeechCohereLabs/cohere-transcribe-03-2026–Word Error Rate (WER): 2.74
speech-recognition-on-voxpopuliCohereLabs/cohere-transcribe-03-2026–Word Error Rate (WER): 5.58
speech-recognition-on-voxpopuli-cleanedCohereLabs/cohere-transcribe-03-2026–Word Error Rate (WER): 5.38

Rank counts only results with a code link.