Voxtral Mini 3B

Voxtral Mini 3B (July 2025) is a compact 3B open-weight speech model built on the Ministral-3B language backbone and released under Apache 2.0. It pairs speech transcription with native audio understanding, so one model can transcribe, translate, summarize, and answer questions about spoken content rather than chaining a recognizer to a separate language model. The model is natively multilingual across widely used languages and supports a dedicated transcription mode alongside built-in question answering and summarization. Function calling lets it trigger downstream actions directly from voice, which suits agentic and tool-using workflows. Its small footprint makes Voxtral Mini 3B a practical choice for edge and on-device deployment of voice agents, conversational interfaces, and cost-sensitive audio applications where efficiency matters.