Mistral AI Launches Voxtral: Open-Source Speech Models

Mistral AI has released Voxtral, its first open-source AI audio model, announced on their website. Voxtral aims to provide businesses with a cost-effective alternative to proprietary speech recognition systems. The models are available in two variants: Voxtral Small, with 24 billion parameters for production environments, and Voxtral Mini, with 3 billion parameters for local and edge deployments.

Voxtral offers advanced speech recognition capabilities, including multilingual support for languages such as English, Spanish, French, and more. It can transcribe up to 30 minutes of audio and understand up to 40 minutes, allowing for real-time actions like API calls. The models are available under the Apache 2.0 license, making them accessible for open use.

Voxtral is designed to bridge the gap between open-source systems with high error rates and expensive proprietary APIs. It provides state-of-the-art accuracy and semantic understanding at less than half the price of comparable solutions. Businesses can integrate Voxtral into applications starting at $0.001 per minute, with options to try the API for free on platforms like Hugging Face.