Mistral AI · 20 Oct 2026
Voxtral 3
Mistral’s four-model audio family for speech synthesis, batch and real-time transcription, diarization and voice agents.
Material model update, access expansion, pricing shift or comparable capability change.
Release is backed by the lab’s own announcement or newsroom post.
Verified source · 1 receiptModel lineage
Where Voxtral 3 sits.
Model lifecycle
Know whether it is ready—and when it is leaving.
Radar tracks the model’s current access stage separately from its original announcement, including official retirement dates and recommended replacements when the lab publishes them.
Available as a preview, beta or experimental release.
- Status since
- 20 Oct 2026
Change Lens
What actually changed.
Normalized from the release note so different labs can be compared by the same fields, not just by launch copy.
Multimodal
Evidence is first-party; comparison labels are normalized by Radar.
Release is backed by the lab’s own announcement or newsroom post.
Verified source · 1 receiptSource receipts
Evidence behind the brief.
The Radar keeps claims traceable: first-party sources, normalized fields, unknowns and the latest verification marker used for this page.
Pricing, Context are not clearly specified in the indexed source note.
Unifies listening, reasoning, speaking and tool use for low-latency voice workflows across more than 50 languages.
This brief links back to the first-party announcement or documentation used by the radar.
Open original source