Mistral Large 4 Goes Open Weight: What It Means (2026)

By PowerfulAI.tools Editorial Team

French AI company Mistral has released the open weights for Mistral Large 4, its flagship model, which it nicknames "Le Chonk". The model first appeared as a paid public preview on 6 October 2026. This week's release is an improved version, and anyone can now download it and run it on their own hardware.

Here is what Mistral announced on 29 October 2026, who it is for, what it costs and what it means if you are not a developer.

Image: Mistral AI

What launched

Mistral Large 4 is a very large model. Mistral says it has 1 trillion parameters, with 52 billion active at a time. It uses a "mixture of experts" design, which means only part of the model works on each request. It accepts multimodal input such as text and images, and Mistral says it is fluent in more than 160 languages, including every official language of the European Union.

It is also a hybrid model: it can answer quickly or reason step by step. Mistral lists three reasoning settings. Low is for simple tasks, Medium is for everyday tasks, and High, the default, is for the hardest problems.

Mistral says the open-weight version improves on the preview, with extra training aimed at coding and cybersecurity. Its own figures include:

• Terminal-Bench 4.0: 41%, up 13 percentage points on the public preview.
• Artificial Analysis Coding Agent Index: 48%, roughly level with GLM-5.3 but using about 47% fewer tokens per task.
• Artificial Analysis Cyber Index: 51%.
• Malicious cyber requests: refused 96.6% of the time on average, which Mistral says is the highest among the open models it measured.

These are Mistral's own published results, not our testing.

Who can use it, and where

This is mainly a release for developers and businesses. Mistral says the model is available through:

• Hugging Face: the open weights can be downloaded and run with common serving tools such as vLLM or SGLang.
• Mistral API and Mistral Studio: Mistral's own developer platform.
• OpenRouter, NVIDIA-hosted endpoints and Microsoft Foundry: third-party platforms, with private deployment options for Azure developers.
• Vibe Code: Mistral's coding tool, where you can try the model on your own code.

Mistral's announcement does not say the model is available in its Le Chat app, so we can't confirm that.

The model is released under the Mistral Ecosystem Open-Weight licence, which Mistral describes as a modified MIT licence. You can use, modify, fine-tune and sell products built on it for free. The exception is businesses that run a "Model-as-a-Service" offering, meaning they sell access to the model itself. They need a separate commercial licence from Mistral.

Mistral also runs a separate cyber access programme. Vetted enterprise partners get preview access to a specialised variant for penetration testing and red-teaming, but only on systems they own or are authorised to test.

What it costs

The open weights are free to download. Running a model of this size yourself needs serious server hardware, so for most people the paid API is the practical option. Mistral's official API pricing is:

• Input: US$0.70 per million tokens
• Cached input: US$0.07 per million tokens
• Output: US$2.10 per million tokens

Prices are in US dollars. A token is a small chunk of text, roughly three-quarters of an English word.

What it means for ordinary users

If you mainly use AI through a chat app, nothing changes for you today. Mistral Large 4 is not a new consumer app, and Mistral hasn't announced it for Le Chat.

The bigger story is that Mistral's flagship model is now free to download. That matters in a few ways:

• More choice in the apps you use: other companies can build Mistral Large 4 into their own products without paying per-use fees to a big US provider.
• Privacy and control: businesses, governments and hospitals can run it on their own servers, so sensitive data doesn't have to leave their systems.

Australian businesses and developers can use it the same way as anyone else. The weights are on Hugging Face, and the API is available through Mistral and the platforms above.

How we checked

We checked every claim in this post against Mistral's own announcements on 30 October 2026: Mistral Large 4 (Le Chonk) is now open weight and Introducing Mistral Large 4. Benchmark results are as published by Mistral. We have not tested the model ourselves.