Hey followers! Today, we’re diving into the exciting world of AI and speech recognition — tech just keeps getting cooler!

French startup Mistral has entered the audio game with Voxtral, its latest open-source AI model designed specifically for businesses. This model promises to bring truly useful speech intelligence into everyday production without breaking the bank or locking users into closed systems.

Voxtral can handle up to 30 minutes of audio for transcription and comprehension, thanks to its robust LLM core, Mistral Small 3.1. It can also process and understand audio up to 40 minutes long, making it possible to ask questions, generate summaries, or execute voice commands on the fly. Plus, it’s multilingual, supporting English, Spanish, French, Portuguese, Hindi, German, Dutch, and Italian.

The company offers two main versions: Voxtral Small with 24 billion parameters for large-scale deployment, comparable to popular solutions like GPT-4o-mini, and Voxtral Mini with 3 billion parameters for smaller, local, or edge usage. There’s even a cost-effective API version—Voxtral Mini Transcribe—that’s optimized for transcription tasks and outperforms OpenAI Whisper at less than half the price.

You can test Voxtral for free on Hugging Face or in Mistral’s chatbot Le Chat. Integration costs start at just $0.001 per minute, making it accessible for many applications. This launch follows Mistral’s recent release of reasoning models that step through complex problems, and the company is actively seeking up to $1 billion in investment to fund its open-source approach.

All in all, Voxtral signifies a major step towards open, affordable, and powerful speech AI for businesses — a game-changer in the AI speech arena!