Mistral AI Releases Voxtral 3: Speech-to-Text, TTS and a Voice Agent SDK

Mistral AI has launched Voxtral 3, a new family of speech models paired with a voice agent SDK built on LiveKit. The release includes three transcription models, one text-to-speech model, and an SDK that supports “listening, reasoning, tool calls and speaking” in a single streaming pipeline. All four models are available in public preview via the Mistral API on EU, US, and global endpoints, or through customer infrastructure. Voxtral Mini Transcribe 3 uses a 4B-parameter model, while Small Transcribe 3’s size wasn’t disclosed. Mistral claims top results on independent speech-to-text rankings and highlights diarization and multilingual batch performance.