**Published:** April 28, 2026 | **Category:** AI Audio Tools
## Introduction
Cohere entry into the automatic speech recognition (ASR) space with Transcribe marks a significant development for the open-source AI community. Released on March 26, 2026, this 2-billion-parameter model immediately claimed the top spot on the Hugging Face Open ASR Leaderboard.
## Performance That Surprises
### Benchmark Results
Cohere Transcribe achieved an average word error rate (WER) of 5.42%—a remarkable achievement that beat:
– **OpenAI Whisper Large v3**: 7.44% WER
– **ElevenLabs Scribe v2**: 5.83% WER
– **Qwen3-ASR-1.7B**: 5.76% WER
### Human Evaluation
In human evaluations, Transcribe was preferred over Whisper Large v3 in 64% of English pairwise comparisons.
## Supported Languages
The initial release supports 14 languages including English, French, Chinese, Arabic, and Japanese.
## Open Source Advantages
### Apache 2.0 License
Transcribe is released under the Apache 2.0 license, one of the most permissive open-source licenses available.
### Self-Hosting Options
Organizations with data privacy requirements can deploy Transcribe entirely on-premises.
## Integration Options
– Hugging Face model hub
– Cohere API
– Model Vault for production deployment
## Conclusion
Cohere Transcribe represents a significant leap forward in open-source speech recognition.
**Rating:** 4.6/5