AI & Models
Cohere launches open source Transcribe voice model
Cohere has launched Transcribe, an open source automatic speech recognition model that claims to outperform several competitors on the Hugging Face leaderboard.
On Thursday, enterprise artificial-intelligence company Cohere launched Transcribe, marking its first entry into the voice model market. Transcribe is an open source automatic speech recognition (ASR) model designed for tasks such as note-taking and speech analysis. Built with 2 billion parameters, the model is designed to run on consumer-grade GPUs, allowing users who want to self-host the system to do so. According to Cohere, Transcribe outperforms several rival models on the Hugging Face Open ASR leaderboard, achieving an average word error rate (WER) of 5.42, which is lower than any other model on the benchmark. In human evaluations assessing transcription accuracy, coherence, and usability, Transcribe achieved an average win rate of 61% over these other models. Cohere compared Transcribe against several competitor models on the leaderboard:
- Zoom Scribe v1
- IBM Granite 4.0 1B
- ElevenLabs Scribe v2
- Qwen3-ASR-1.7B Speech
The model currently supports 14 languages: English, French, German, Italian, Spanish, Portuguese, Greek, Dutch, Polish, Chinese, Japanese, Korean, Vietnamese, and Arabic. In terms of processing speed, Transcribe can process 525 minutes of audio in a single minute, which is high for its class of model. However, the model has notable limitations. When transcribing Portuguese, German, and Spanish, Transcribe fell behind its rivals.
Cohere is planning to integrate Transcribe into North, its enterprise agent orchestration platform. The company is also making the model available for free through its API, and it will be available on Model Vault, Cohere’s managed inference platform. The launch comes amid growing demand for note-taking and dictation applications like Granola and Wispr Flow. Financially, Cohere was reportedly generating annual recurring revenue (ARR) of $240 million in 2025. The startup’s CEO, Aidan Gomez, has indicated that the company may go public “soon”.
Why it matters
Cohere’s move into open source speech recognition signals a push to capture the growing demand for automated note-taking and dictation tools, directly challenging established players in the enterprise AI space.