The model family supports more than 50 languages and includes two batch transcription models and one real-time streaming model. Voxtral 3 TTS ranks 12th in both the Provider Voice and Controlled Voice arenas and scored 84.4% on the Pronunciation Robustness benchmark. These tools are part of a voice agents API capable of automating billing inquiries, appointment booking, and identity verification.
Tuesday, October 20, 2026
Story
Mistral Voxtral 3 Takes No. 3 Spot on Speech to Text Leaderboard
AI8 hours ago
Mistral has deployed its Voxtral 3 speech models to OpenRouter with endpoints hosted in the European Union and zero-data-retention. The Voxtral Small Transcribe 3 model achieved a 2.1% AA-WER Index, placing it at number 3 on the Artificial Analysis Speech to Text leaderboard, ahead of ElevenLabs Scribe v2 at 2.2% and Grok Voice Transcribe 2.0 at 2.3%. The accompanying Voxtral 3 TTS model generates 112.3 characters per second—surpassing Eleven v4 Turbo's 78.1—and costs $13 per 1 million characters, undercutting Gemini 3.8 Flash TTS at $16.5.
Looking for more context or analysis?
Tilt chat provides reliable, tailored insights, alerts, and automations.