Skip to main content

Which models are used for voice transcription?

Updated this week

The accuracy and speed of the Voice mode feature in le Chat are made possible by a specialized Speech-to-Text (STT) model developed by Mistral AI: voxtral-mini-2507

This model is specifically optimized for fast and accurate transcription of spoken language across various contexts and languages, including:

  • English

  • Français (French)

  • Deutsch (German)

  • Español (Spanish)

  • Italiano (Italian)

  • Nederlands (Dutch)

  • Português (Portuguese)

  • हिन्दी (Hindi)​

By using our own state-of-the-art model, we aim to provide a high-quality and seamless voice interaction experience for all users.

🔎 For more details on the capabilities and architecture of this model, you can read the official blog post announcement.

Did this answer your question?