Mistral presents Voxtral 3, a voice agent SDK for 50+ languages
🎧 Listen to this story
Mistral presents Voxtral 3, a family of speech-to-text and text-to-speech models alongside an SDK that combines listening, reasoning, tool use and speaking in a real-time streaming pipeline. The company says Voxtral Small Transcribe 3 transcribes more than 50 languages, identifies who spoke and detects when speech has ended; Voxtral Mini TTS 3 offers voices that adapt to context, supporting more than 35 languages and 50 dialects.
The SDK is designed for contact-center agents, virtual assistants and copilots that perform tasks in connected systems during phone or web conversations. Mistral says its Pro Voice feature can reproduce a speaker’s character, emotional range and speaking style from 30 minutes of audio. Voxtral 3 is available through Mistral’s EU-hosted API or for deployment on a customer’s infrastructure, allowing control over audio processing and retention.
Why it matters · editorial interpretation
By combining transcription, reasoning, tool use and speech synthesis in a streaming pipeline, the SDK could simplify the creation of voice agents connected to other systems. Its EU-hosted API or deployment on a customer’s infrastructure also gives teams a choice over how audio is processed and retained.
Sources
APIsAgents