Mistral AI Releases Voxtral: The Worlds Best (and Open) Speech Recognition Models

Mistral AI has introduced Voxtral, a family of open-weight modelsincluding Voxtral-Small-24B and Voxtral-Mini-3Bthat combine automatic speech recognition (ASR) with natural language understanding, enabling seamless processing of both audio and text inputs. Built on Mistrals language modeling framework and supporting a 32,000-token context window, these models facilitate tasks such as transcription, summarization, and question answering, with the ability to handle audio durations of up to 30-40 minutes without segmentation, making them highly suitable for enterprise and multimedia applications. The models