TechPrufer
Get started
Mistral AI
Mistral AI logoMistral AI · Feature trail

Voxtral Speech Understanding Models

3 posts · Jul 2025 – Mar 2026

What this trail covers

Mistral AI releases Voxtral, state-of-the-art speech understanding models available in 24B and 3B variants under Apache 2.0, and via API. Voxtral offers advanced transcription, built-in Q&A and summarization, multilingual support, and function-calling directly from voice. Benchmarks show superior performance to Whisper, GPT-4o mini, and Gemini 2.5 Flash in transcription and audio understanding tasks. Enterprise features include private deployment, domain-specific fine-tuning, and advanced context support. Future updates will include speaker segmentation, emotion detection, word-level timestamps, and non-speech audio recognition.

All Mistral AI trails
  1. Speaking of Voxtral | Mistral AIThis post details the release of Voxtral TTS, a 4B-parameter text-to-speech model. It highlights its realistic and emotionally expressive speech generation in 9 languages, low latency, and easy voice adaptation. The model's architecture is described as a transformer-based, autoregressive, flow-matching model built on Ministral 3B, comprising a transformer decoder backbone, a flow-matching acoustic transformer, and a neural audio codec. Performance comparisons with ElevenLabs are presented, showcasing superior naturalness and similar Time-to-First-Audio (TTFA) while supporting emotion-steering. The post also details zero-shot cross-lingual voice adaptation capabilities and its application in cascaded speech-to-speech translation.
  2. Voxtral transcribes at the speed of sound. | Mistral AIThis post details the release of Voxtral Transcribe 2, introducing Voxtral Mini Transcribe V2 and Voxtral Realtime. Voxtral Realtime features a novel streaming architecture for sub-200ms latency, enabling live applications, and is released as open-weights under Apache 2.0. Voxtral Mini Transcribe V2 offers improved batch transcription quality, diarization, context biasing, and word-level timestamps across 13 languages, with a focus on cost-efficiency and noise robustness. The models support up to 3-hour audio inputs and GDPR/HIPAA compliance.
  3. Voxtral | Mistral AIThis post introduces the Voxtral models, detailing their architecture (24B and 3B variants, based on Mistral Small 3.1 for text capabilities), performance benchmarks against leading models (Whisper, GPT-4o mini, Gemini 2.5 Flash, ElevenLabs Scribe) across English and multilingual transcription and audio understanding tasks, and key features like long-form context (32k for transcription, 40 min for understanding), built-in Q&A/summarization, multilingual fluency, and function-calling. It also outlines enterprise offerings for private deployment and fine-tuning, and previews upcoming features like speaker segmentation and word-level timestamps.