Mistral AI launches Voxtral 3 speech models and a real-time agent SDK

Voxtral 3

Mistral AI released the Voxtral 3 family of speech transcription and synthesis models, plus a real-time agent SDK that combines listening, reasoning, tool calls and speech output. Transcription covers more than 50 languages with under 500ms latency in real time. Synthesis covers more than 35 languages and 50 dialects, adjusting emotion, pace and emphasis from context.

Why it matters: Mistral wires transcription, reasoning, tool calls and speech synthesis into one real-time pipeline, and each stage's model can be swapped without rewriting the voice agent.