AI-powered features are becoming increasingly widespread, with many of them relying heavily on voice commands. In line with this trend, French startup Mistral has unveiled its first open-source voice AI, named Voxtral, aiming to compete with major players in the field.
Mistral describes Voxtral as the first voice AI model capable of delivering “truly practical speech intelligence” in real-world environments. Thanks to this model, developers no longer have to choose between cheap, underperforming systems and powerful but expensive proprietary alternatives.
According to Mistral, businesses can deploy the Voxtral model at less than half the cost of comparable models. This new French model is capable of transcribing up to 30 minutes of audio content.
Using the Mistral Small 3.1 language model, Voxtral can comprehend up to 40 minutes of audio and allows users to ask questions about the content, generate summaries, and issue voice commands. The model supports English, Spanish, French, Portuguese, Hindi, German, Dutch, and Italian.
There are two versions of the Voxtral AI model:
- Voxtral Small with 24 billion parameters, designed to compete with models like GPT-4o-mini, ElevenLabs Scribe, and Gemini 2.5 Flash.
- Voxtral Mini with 3 billion parameters, optimized for local deployment.
Additionally, Voxtral Mini has a more affordable and faster variant called Voxtral Mini Transcribe, built specifically for transcription tasks. According to Mistral, this version outperforms OpenAI Whisper at less than half the cost.
Users can test the model for free via Hugging Face or Mistral’s dedicated chatbot Le Chat. Integration of Voxtral into applications starts at $0.001 per minute.
Mistral is considered one of Europe’s leading AI companies and is reportedly seeking to raise up to $1 billion in funding from firms such as MGX Abu Dhabi.
