Europe's New AI Mistral Aims to Beat US and China

French AI lab Mistral just unveiled Mistral Large 4, a giant AI model with 1 trillion parameters (the dials that control its intelligence). Nicknamed 'Le Chonk,' it's a response to the growing split between closed AI from companies like OpenAI and open AI from China. Mistral pitches it as a 'third way' — European, powerful, and eventually open for anyone to inspect and use.

For now, you can only access ML4 through a guarded online endpoint, but in three weeks Mistral plans to release the full model weights. That means anyone can download, study, and modify it — a big deal for security and trust. Mistral says it trained ML4 on just 4,000 Nvidia GPUs, two to three times less than Chinese rivals and far less than closed competitors. This efficiency could make AI cheaper and more accessible.

The model is multimodal, meaning it can understand text, images, and more. It's optimized for cybersecurity, finance, and chip design — fields where Mistral's backers, like ASML and Samsung, have deep interests. While official test scores aren't out yet, Mistral hopes ML4 will be the best open model outside China, and even beat closed models in specific tasks.

So what? If ML4 lives up to its promise, it could give businesses and governments a safer, auditable alternative to black-box AI. It might also pressure other AI makers to be more open and efficient. For everyday people, that could mean better privacy, lower costs, and more control over the AI that increasingly runs our world.