Voxtral Speech 3

Spec

Attribute	Value
Developer	Mistral AI
Released	2026-10-20
Announced	2026-10-20
Context window	unknown
Pricing	$13/1M chars
License	unknown
Availability	Mistral API; contact Mistral for private deployment
Specifications follow the launch announcement; context window and license terms are not established by that source. (source)

Release Date

Announced and available through the API on 2026-10-20. Speech generation in batch and realtime, also named Voxtral Mini TTS 3.0 in the announcement; supports 35+ languages and 50+ dialects. (source)

Benchmarks

Voice Code Bench uses 300 English prompts, 2 voices per system and gemini-3.8-flash as judge. The announcement also reports internal pronunciation, multilingual and expressiveness evaluations. These are vendor-presented evaluations; numerical chart coordinates were not reliably extracted. (source)

Use Cases

Context-sensitive speech with explicit emotion, pacing and emphasis controls; Pro Voice adaptation uses 30+ minutes of sampled audio. (source)

Compared To

The Voxtral voice pipeline uses separate transcription, reasoning and speech-generation models. Its reasoning stage can be replaced independently; family-level voice-agent claims are not this model's standalone benchmark. See Agents (LLM Agents). (source)

Sources

• Launch announcement and immutable capture; linked model documentation and model weights were not reviewed.