<ul data-eligibleForWebStory="false"><li>Mistral released an open-sourced voice model called Voxtral that offers advanced features like summarization and speech-triggered functions.</li><li>Voxtral comes in a 24B parameter version for scale applications and a 3B variant for local and edge use cases.</li><li>Mistral aims to bridge the gap between proprietary speech recognition models and open-source versions with Voxtral, providing accurate transcription, semantic understanding, multilingual fluency, and flexible deployment at half the price of comparable APIs.</li><li>Voxtral outperforms existing voice models, offering fewer word errors compared to other models and competitive performance in audio understanding tasks. It will be available through Mistral's API at $0.001 per minute.</li></ul>

Mistral’s Voxtral goes beyond transcription with summarization, speech-triggered functions

Discover more