Mistral launches Voxtral 3 speech models from $0.0035 a minute

Mistral AI released Voxtral 3 on October 20, 2026, a family of four speech models plus a voice agent SDK built on LiveKit. The cheapest model, Voxtral Mini Transcribe 3, costs $0.0035 per minute of audio for batch transcription. The family also adds text-to-speech at $13 per 1 million characters, and a voice agent API that runs listening, reasoning and speech as one streaming pipeline from $0.04 a minute.

What it costs

Model	What it does	Price
Voxtral Mini Transcribe 3	Batch speech-to-text (4B parameters)	$0.0035/min
Voxtral Small Transcribe 3	Most accurate speech-to-text	$0.0040/min
Voxtral Mini Transcribe Realtime 3	Live speech-to-text for agents	$0.008/min
Voxtral Mini TTS 3	Text-to-speech, batch and realtime	$13 per 1M characters
Voice agent API with Mistral Small 4	Full voice agent	$0.04/min
Voice agent API with a larger model	Full voice agent	$0.08/min

Mistral’s post does not mention a free tier.

What changed

Mistral says transcription covers more than 50 languages, including the 24 official EU languages. The post gives two different figures for realtime transcription, 35+ and 40+ languages, and does not say which is current. Text-to-speech covers 35+ languages and 50+ dialects, with voice presets in 18 languages.

The models are on the Mistral API through EU, US and global endpoints, and can be tried in Mistral’s Studio playgrounds. Mistral also offers self-deployment and a full on-premises or private-cloud stack through its sales team. The post does not say whether the weights are public or under what license, and it does not mention retiring the earlier Voxtral 2 models.

Who it affects

Developers building transcription, captioning or phone and voice agents. At $0.0035 a minute, an hour of batch transcription costs about $0.21. The voice agent API bundles speech-to-text, a Mistral language model and speech output into a single per-minute price, which makes costs easier to estimate than paying for three services separately. To compare voice options, see our AI voice comparison.