Mistral releases Voxtral 3 voice agent SDK for real-time speech
10/20/2026 — 10/21, 02:19·3 sources (all)·6 reports (all)·In the 2026-10-21 briefing
Key takeaways
- Mistral AI released Voxtral 3, a voice agent SDK built for real-time speech that handles accurate transcription in more than 50 languages, e…
Background & analysis
On October 20, 2026, Mistral AI released Voxtral 3, which it describes as a voice agent SDK built for real-time speech. According to the company's announcement, the SDK offers accurate transcription in more than 50 languages, emotional TTS, and deep tool use. Voxtral 3 can be used either through EU-hosted deployment or by self-deploying it. The available information stops at the launch itself; no pricing, availability details, or further roadmap were disclosed in the material.
AI-generated from 1 reports · updated 3 hours ago
Latest turnMistral AI has released Voxtral 3, a family of models for real-time voice agents that combines speech-to-text, text-to-speech and a voice agent SDK. It supports more than 50 languages, lets agents call external tools mid-conversation, and can run on your own infrastructure or through EU-hosted APIs.
Related AI tools
- Heat index
- 270
- All sources
- 3
- All reports
- 6
- First seen
- 7 hours ago
Reports on this story headlines open the original
Mistral AI has released Voxtral 3, a family of models for real-time voice agents that combines speech-to-text, text-to-speech and a voice agent SDK. It supports more than 50 languages, lets agents call external tools mid-conversation, and can run on your own infrastructure or through EU-hosted APIs.
The DecoderAI score 67Voxtral 3 TTS generates 112.3 characters per second, ahead of Eleven v4 Turbo at 78.1, Simba 3.2 at 77.8 and Gemini 3.8 Flash TTS at 55.3, according to Artificial Analysis. It is also priced at $13 per 1M characters, below Gemini 3.8 Flash TTS at $16.5 per 1M.
X：Artificial Analysis (@artificialanlys)First-partyAI score 40Voxtral Mini Transcribe Realtime 3 ships with two latency settings, both landing in the top half of Artificial Analysis's streaming speech-to-text leaderboard. At the 960 ms setting it posts a 3.3% final-transcript AA-WER, ranking sixth of 46 models while returning a final transcript 1.1 seconds after speech ends.
X：Artificial Analysis (@artificialanlys)First-partyAI score 40Voxtral Small Transcribe 3 ranks third on Artificial Analysis's speech-to-text leaderboard with a 2.1% AA-WER index, ahead of ElevenLabs Scribe v2 at 2.2% and Grok Voice Transcribe 2.0 at 2.3%. Voxtral Mini Transcribe 3 lands fifth at 2.2%.
X：Artificial Analysis (@artificialanlys)First-partyAI score 40Mistral has released Voxtral 3, a new generation of speech models covering both text-to-speech and transcription. According to Artificial Analysis, Voxtral Small Transcribe 3 ranks third on its speech-to-text leaderboard, while Voxtral 3 TTS enters the top 12 in both text-to-speech arenas.
X：Artificial Analysis (@artificialanlys)First-partyAI score 56
Mistral AI released Voxtral 3, a voice agent SDK for real-time speech that covers accurate transcription, emotional TTS and deep tool use across more than 50 languages. It is available EU-hosted or self-deployed.
Mistral AIFirst-partyAI score 61
Other stories people are talking about
- Heat index 345Four Gemini 4 Argon agents beat Minecraft in 27 hours7 sources
- Heat index 79Same photo billed 1,120 tokens on Gemini vs 14,364 on GPT-5.61 sources
- Heat index 87Mistral releases Timestral time-series model family2 sources
- Heat index 1044RisingDHS orders federal civilian agencies to purge DeepSeek tech16 sources
- Heat index 892RisingDeepSeek open-sources DeepInfer large-scale inference framework11 sources
How is heat calculated?About the methodHide
Heat counts how many independent sources covered a story in the last 48 hours: one source counts once no matter how many posts it published, decaying with a 24-hour half-life. What ranks first is what many people are talking about.
This page aggregates public feeds. Headlines and summaries are machine-organized and remain the property of the original authors; verify important facts at the source. If you believe a headline or summary infringes your rights, email the contact address in the footer with the page URL and basis of your claim, and we will remove or replace it after review.
- Surge
- Discussion rising fast
- New
- First report within 6 hours
- Rising
- Still gathering discussion