Mistral Large 4 is now open: what it means for AI users

Short answer

On October 29, 2026, French AI lab Mistral released the open weights of Mistral Large 4, its biggest model, nicknamed "Le Chonk". Anyone can now download it, run it on their own servers and adapt it, under a modified MIT license. You won't run it on a laptop, but it gives apps and companies a strong, European-trained model they can host privately.

What happened

Mistral first showed Large 4 in a public preview earlier in October. After about three more weeks of training, it published the improved version's weights on Hugging Face, and the model is also available through Mistral's own API and services such as OpenRouter.

According to Mistral's model card, Large 4 has about 1 trillion parameters in total, but only a small part of them works on each word (a "mixture of experts" design). It reads text and images, has a context window of 1 million tokens, and is fluent in more than 160 languages, including every official EU language. It also has three reasoning modes, low, medium and high, so the same model can give a quick reply or think through a harder task.

Mistral says Large 4 is the best open-weight model for coding and matches the leading open model on the Artificial Analysis Coding Agent Index while using 47% fewer tokens per task. Those are the company's and benchmark makers' numbers, not independent everyday tests. The license lets people use, modify and sell products built on the model for free; only businesses that rent the model out as a service need a separate license from Mistral.

Why people are talking about it

The release went straight to the top of tech discussion: on Hacker News, Mistral's announcement collected well over 1,000 points in a day, and TechCrunch, ZDNet and Hugging Face all covered it the same day. Open models that come close to the big closed ones are rare, and this one was trained from scratch in Europe.

What it means if you use AI

• More choice behind your apps. Open weights mean more chatbots and tools can run a top-tier model without sending your messages to one big provider.
• Privacy options. Companies and public institutions can keep data on their own servers while still using a strong model.
• Not a home model. A model this size needs far more memory than a normal laptop or phone has, so most people will meet it inside apps and services.
• Better for non-English speakers. Support for 160+ languages helps people who chat with AI in their own language.

Quick checklist

1. Want to try it? Use a hosted option such as Mistral's own products or OpenRouter rather than downloading it.
2. Check where an app runs its model if privacy matters to you.
3. Treat benchmark rankings as a guide, and test it on questions you know the answer to.
4. Keep private details out of prompts in any AI tool, open or closed.

FAQ

Is Mistral Large 4 free? The weights are free to download and use under Mistral's license; using it through hosted services can cost money.

Does open mean less safe? Mistral says it added guardrails and reports high refusal rates for harmful cyber requests, and it says it will keep updating them.

Facts come from Mistral's announcement and model card of October 29, 2026, with reporting from TechCrunch, ZDNet and the Hugging Face blog. Benchmark figures are reported by Mistral and Artificial Analysis.

Related reading:

• Gemini 4 Carbon: Google's new #1 AI model, coming soon
• ChatGPT vs Claude vs Gemini for everyday life
• Is ChatGPT safe? What happens to what you type

Cover image: Arthur Mensch, co-founder and CEO of Mistral AI, at Techarena 2026. Photo: Jan2342342423, Wikimedia Commons, CC0 (public domain).