Capturing data…
Capturing data…
THU, 22 OCT · 89 ITEMS
"Mistral launches Voxtral 3 AI model" — Hacker News (best) · AI & Technology
Mistral has released 3, a family of speech models and a voice agent enabling real-time voice interactions. It includes transcription models supporting over 50 languages, emotional text-to-speech, and tool-use capabilities, making it a full-stack solution for building voice agents that listen, reason, and speak.
The claim is directly from Mistral's official blog post (a primary source) and corroborated by multiple tech outlets. As a corporate announcement, the factual claims are authoritative, though specific benchmark rankings and language support figures come from the company.
Genuinely new: announced approximately one day ago (circa October 20, 2026), with no indication of recirculated older material.