Mistral releases a new open-source model for speech generation Mistral’s new speech model can run on a smartwatch or a smartphone.