Otto365 AI Brief #022
Mistral: Voxtral voice agents could answer your phones
What happened
Could Mistral's new Voxtral voice agents answer your phones and book appointments? Mistral AI announced Voxtral 3, speech models plus a voice agent SDK, on the 20th of October 2026. It listens, reasons, uses your tools and talks back, over phone or web.1
Three things to know
- Hears 50+ languages First, Voxtral Transcribe 3 turns speech into text in over 50 languages. It works out who said what, takes your technical terms, and spots when a caller has finished speaking.1
- Speaks 35+ languages Second, Voxtral Speech 3 replies in 35 languages and over 50 dialects, matching emotion and pace to the conversation. It costs 13 dollars per million characters, and Pro Voice clones a voice from 30 minutes of audio.1
- Agent takes the call Third, the agent SDK: on a live call it can verify who is calling, check a bill or book an appointment. Developers run it on Mistral's EU-hosted API or their own servers.1
Why it matters to you
Otto365's take (opinion)
So, my take. Useful for bookings and multilingual callers, but this is a developer kit most UK small firms would use through a supplier. Ask where audio is processed and retained, and keep a human route.1
Do this week
Try this: list the routine calls a voice agent could take.
Sources
- Mistral AI: Voxtral 3: Voice Agent SDK & Speech Models in 50+ Languages (opens in a new tab) Primary source, accessed 21 Oct 2026
The quotes we checked (31)
we introduce Voxtral 3: a new family of state-of-the-art speech-to-text and text-to-speech models, alongside a voice agent SDK
Mistral AI · source 1
brings listening, thinking, tool calling, and speaking into a single real-time streaming pipeline.
Mistral AI · source 1
prices from vendor pricing pages; generated 2026-10-20.
Mistral AI · source 1
Voxtral 3 is available today through the Mistral API, hosted through our European, US, and global endpoints.
Mistral AI · source 1
Agents can retrieve information, interact with business systems, and complete tasks during live conversations over phone or web.
Mistral AI · source 1
Accurate transcription across 50+ languages. The speech-to-text model, Voxtral Small Transcribe 3, is state-of-the-art in multilingual transcription
Mistral AI · source 1
Voxtral Transcribe 3 represents a significant milestone across transcription, diarization (who spoke when) and context biasing capabilities.
Mistral AI · source 1
identifying who said what (diarization), using keyphrases to improve its accuracy (context-biasing).
Mistral AI · source 1
It detects when someone has finished speaking so agents can respond without delay (end-of-turn detection).
Mistral AI · source 1
key terms and free-form text can be given to bias the transcription and improve recall on technical terms
Mistral AI · source 1
It outperforms Whisper, Scribe v2 and Gemini 3.5 Transcribe.
Mistral AI · source 1
The supported list of languages is expanded to 50, including 24 official EU languages and many low-resource Asian and Indic languages.
Mistral AI · source 1
It achieves high-quality transcriptions at sub-500ms latency in over 35 languages, including with diarization and context biasing.
Mistral AI · source 1
Voxtral Mini TTS 3.0, available in batch and realtime, achieves state-of-the-art expressiveness and instruction following for 35+ languages and 50+ dialects.
Mistral AI · source 1
Speech sounds local in 35 languages and over 50 dialects.
Mistral AI · source 1
new voices that adjust their emotion, pace, and emphasis based on conversational context
Mistral AI · source 1
Voxtral TTS natively interprets a prompt’s text context to determine the emotion and voice style
Mistral AI · source 1
At $13 per 1M characters, Voxtral Speech 3 costs less to run than comparable models from OpenAI TTS-1, Deepgram, ElevenLabs Flash and Gemini Flash.
Mistral AI · source 1
Voxtral’s Pro Voice adaptation is able to reproduce a speaker’s distinctive character, emotional range, and speaking style across languages, using just 30 minutes of audio.
Mistral AI · source 1
How about the Emirates flight at 21,000pm on 10 over 10, returning 10 over 25 at 22,000pm for £1,881?
Mistral AI · source 1
How about the Emirates flight at 21pm on October 10th, returning October 25th at 22pm for £1,881?
Mistral AI · source 1
Voxtral 3 introduces a voice agent SDK, built on LiveKit, that runs speech in, reasoning, and speech out as one streaming pipeline.
Mistral AI · source 1
During a call, the agent can verify a caller's identity, check a bill, or book an appointment, using function calling, web search, and retrieval over your own systems.
Mistral AI · source 1
It answers and places calls over SIP, runs in browsers and apps over WebRTC, and connects to the telephony, IVR, and contact-center software you already use.
Mistral AI · source 1
No stage waits for the one before it, so each turn completes in under one second, and the agent stops talking when the caller cuts in.
Mistral AI · source 1
Available today through Mistral's EU-hosted API or for self-deployment on your own infrastructure, it gives organizations control over where audio is processed, how data is handled, and whether conversations are retained.
Mistral AI · source 1
Keep sensitive conversations within your environment and control whether any audio is retained.
Mistral AI · source 1
Hi, my doctor gave me a referral letter in French. Can you book it for me?
Mistral AI · source 1
Sure. What's the reference number?
Mistral AI · source 1
Your doctor is asking a dermatologist, Dr Martin, to see you. The next slot is Tuesday at 5:45 pm. Shall I book it?
Mistral AI · source 1
Done. Bring the letter, your carte Vitale, and I'll text you the details in English.
Mistral AI · source 1
Transcript
Full transcript
Could Mistral's new Voxtral voice agents answer your phones and book appointments?
Play from 0:04 Mistral AI announced Voxtral 3, speech models plus a voice agent SDK, on the 20th of October 2026. It listens, reasons, uses your tools and talks back, over phone or web.
Play from 0:19 First, Voxtral Transcribe 3 turns speech into text in over 50 languages. It works out who said what, takes your technical terms, and spots when a caller has finished speaking.
Play from 0:30 Second, Voxtral Speech 3 replies in 35 languages and over 50 dialects, matching emotion and pace to the conversation. It costs 13 dollars per million characters, and Pro Voice clones a voice from 30 minutes of audio.
Play from 0:46 Third, the agent SDK: on a live call it can verify who is calling, check a bill or book an appointment. Developers run it on Mistral's EU-hosted API or their own servers.
Play from 0:59 So, my take. Useful for bookings and multilingual callers, but this is a developer kit most UK small firms would use through a supplier. Ask where audio is processed and retained, and keep a human route.
Play from 1:13 Try this: list the routine calls a voice agent could take.
I'm Otto365.
Follow the Brief.
On screen
- Play from 0:00
- Otto365 AI Brief #022, Mistral, 20 Oct 2026
- Mistral's voice agents could answer your phones
- Customer calls
- Voice agent
- Books a slot
- Confirmed
- Play from 0:04
- Voxtral 3: voice agents
- Announced 20 Oct 2026
- Listens
- Reasons
- Calls tools
- Speaks
- Source: Mistral AI, 20 Oct 2026
- Play from 0:19
- Hears 50+ languages
- Caller speaks
- Who said what
- Your key terms
- Spots turn end
- 50 languages, 24 official EU
- Realtime: under 500ms
- Mistral says: beats Whisper
- Play from 0:30
- Speaks 35+ languages
- Your text
- Reads context
- Emotion and pace
- 35+ languages
- Voxtral 1: Says 10 over 10
- Voxtral 3: Says dates right
- Play from 0:46
- Agent takes the call
- Hi, my doctor gave me a referral letter in French. Can you book it for me?
- Sure. What's the reference number?
- A dermatologist can see you. Shall I book it?
- Done. I'll text you the details in English.
- Checks the referral
- Books the appointment
- Phone via SIP, web via WebRTC
- Stops when the caller cuts in
- EU-hosted API or own servers
- Play from 0:59
- Ask where audio is kept
- Gain: Bookings by phone, 50+ languages
- Watch: Audio retention, Developer kit
- EU-hosted API or own servers
- Control if audio is retained
- Keep a human route
- Play from 1:13
- List calls to hand over
- Voiced by AI. Checked by humans. Apps 365 sells AI and Microsoft 365 services. Not endorsed by companies named.
Voiced by AI. Checked by humans. Apps 365 sells AI and Microsoft 365 services. Not endorsed by companies named. Report an error.