Mistral Launches Open-Source Voxtral Audio AI

French startup Mistral has unveiled its first audio model suite, Voxtral, designed specifically for business applications and available as an open-source solution. Voxtral provides an affordable and efficient alternative among audio AI tools, capable of transcribing audio clips up to 30 minutes in length. Powered by the large Mistral Small 3.1 model, Voxtral can process recordings of up to 40 minutes, automatically generating summaries, answering questions related to the audio content, and converting voice commands into instant actions.

The system supports multiple languages, including English, French, Spanish, Portuguese, Hindi, German, Dutch, and Italian, ensuring high-accuracy transcription and deep content understanding. Voxtral is offered in two main versions: Voxtral Small, with 24 billion parameters, targeting large-scale business use cases and competing with solutions like ElevenLabs Scribe and GPT-4o-mini; and Voxtral Mini, a lightweight 3-billion-parameter model optimized for local and edge deployments. There is also a rapid API-only version called Voxtral Mini Transcribe focused purely on transcription tasks.

Users can freely test Voxtral’s capabilities via the Hugging Face API or Mistral’s chatbot “Le Chat.” Pricing starts as low as $0.001 per minute, making it an accessible and attractive choice for companies seeking powerful open-source audio analytics integrated into their workflows.

Time is the most valuable resource for business today. Almost half of it is wasted on routine tasks. Your employees are constantly forced to perform monotonous tasks that are difficult to classify as important and specialized. You can leave everything as it is by hiring additional employees, or you can automate most of the business processes using the ApiX-Drive online connector to get rid of unnecessary time and money expenses once and for all. The choice is yours!