Cohere Launches Transcribe, an Open-Source Speech Recognition Model with 2 Billion Parameters

Cohere has launched Transcribe, an open-source speech recognition model with 2 billion parameters, optimized for self-hosting on consumer GPUs and supporting 14 languages.

Cohere Launches Transcribe, an Open-Source Speech Recognition Model with 2 Billion Parameters
Cohere Transcribe speech recognition model

On March 26, 2026, Cohere unveiled Transcribe, its first open-source speech recognition model, designed for tasks such as meeting transcription and speech analysis. With just 2 billion parameters, Transcribe is optimized for self-hosting on consumer-grade GPUs, making it accessible to a wide range of users and applications.

Key Features of Transcribe

Transcribe supports 14 languages: English, French, German, Italian, Spanish, Portuguese, Greek, Dutch, Polish, Chinese, Japanese, Korean, Vietnamese and Arabic. This broad language support makes it versatile for global applications. In addition, the model achieved an average word error rate (WER) of 5.42%, ranking first on the Hugging Face Open ASR leaderboard and outperforming other open-source and closed models. It can process 525 minutes of audio in one minute, demonstrating high efficiency.

Implications for the Market and Businesses

The launch of Transcribe is a significant challenge for existing speech recognition models, offering a high-performance open-source alternative. Businesses can benefit from greater control over sensitive data, reducing their reliance on external cloud services. In addition, optimization for self-hosting on consumer hardware makes Transcribe a cost-effective and scalable solution for internal transcription needs.

Future Outlook

Cohere plans to integrate Transcribe into its orchestration platform for enterprise agents, North, and to make it available via API and on its managed inference platform, Model Vault. This signals a continued commitment to evolving speech recognition solutions and delivering advanced tools for businesses.

Takeaway

  • Innovation in speech recognition: Transcribe offers an open-source model with superior performance, challenging existing market leaders.
  • Data control: Companies can self-host the model, ensuring greater privacy and security for sensitive data.
  • Efficiency and scalability: Optimization for consumer GPUs and high processing speed make Transcribe a practical solution for everyday business needs.
  • Future integration: The planned integration with Cohere's North platform and availability via API and Model Vault signal a continued commitment to evolving speech recognition solutions.

For more details, you can read the original article on TechCrunch.