Enterprise AI company Cohere has launched an open-source automatic speech recognition (ASR) model called Transcribe, TechCrunch reported on March 26.
Developed with 200 million parameters, the model can run on consumer-grade GPUs and supports 14 languages including English, French, German and Italian.
The company said Transcribe delivers better performance than Zoom Scribe v1, IBM Granite 4.0 1B, ElevenLabs Scribe v2 and Qwen3-ASR-1.7B speech models. It recorded an average word error rate (WER) of 5.42 on the Hugging Face Open ASR leaderboard, beating rival models. But TechCrunch reported that its performance in Portuguese, German and Spanish was relatively lower.
It also highlighted that Transcribe can process 525 minutes of audio in 1 minute. Cohere plans to integrate the model into its enterprise AI platform North and offer it free via an API.
Chi-gyu Hwang
delight@d-today.co.kr