Voxtral Transcribe 3

Voxtral Transcribe 3 is a multilingual speech‑to‑text model part of Mistral’s Voxtral 3 family. It achieves state‑of‑the‑art accuracy across 50+ languages, supports diarization (speaker‑level attribution), detects end‑of‑turn speech, and uses context‑biasing to improve transcription of technical terms. The small version outperforms competitors such as Whisper, Scribe v2, and Gemini 3.5 Transcribe, and the Mini variant runs at 4 B parameters, enabling fast, accurate transcriptions at $0.0035 per minute. Voxtral 3 includes the model Voxtral Transcribe 3 as a multilingual transcription capability· Graphiti 生成的实体摘要