New models
Mistral launches Voxtral 3 voice agent models for real-time conversations
Mistral AI launched Voxtral 3 on 20 October, with transcription priced from $0.0035 per minute, according to The Decoder, which doesn't name the currency. The package combines speech-to-text, text-to-speech and a LiveKit-based voice agent SDK, a toolkit for building agents, in one streaming pipeline. Transcription covers more than 50 languages, and speech output covers more than 35 languages and 50 dialects. Mistral says it can run on your own infrastructure or through its EU-hosted API, and agents can book appointments or check invoices mid-call.
Why it matters: If you take phone or web bookings, a voice agent could handle simple requests. Mistral's hosting is in the EU or on your own systems, and Australian availability and pricing aren't announced, so check where call data is processed.
Source: The Decoder · Also reported by Mistral AI news