Mistral releases Large 4 open-weight AI model, nicknamed Le Chonk

French AI company Mistral has released the weights of Mistral Large 4 (ML4), nicknamed Le Chonk. The company says the model improves on coding and cyber defence. Mistral claims it is the best open agentic coding model in the world.

The release follows a preview on 6 October. Euronews reported that Mistral spent three weeks reinforcing the model's learning and testing it with customers and partners. Mistral says the Terminal-Bench 4.0 score rose from 28% to 41% after the extra training, and that the release uses 47% fewer tokens per task on average than GLM-5.3.

According to Mistral, the model has 1.05 trillion parameters in total and activates around 52 billion. The Hugging Face blog describes it as a multimodal, multilingual Mixture-of-Experts model that takes text and images as input and generates text, with a context window of 1M tokens. Mistral says it was trained on 3,800 Nvidia Grace Blackwell GPUs in its European data centres and offers three reasoning modes: low, medium and high.

Mistral says the model matches or surpasses models from DeepSeek and Z.ai, according to Wired. Hugging Face reported that the model ranks fifth among open-weight models on the Artificial Analysis Intelligence index as of 29 October 2026. It also reported scores of 61.7% on DeepSWE v1.1 and 28.3% on Terminal-Bench 4, and said the model solves 93% of a suite of 40 Cybench challenges.

On security, Mistral reports that the model scores 82% on the Artificial Analysis Cyber Index test of reproducing and patching vulnerabilities. It says several closed frontier models score near zero because they refuse to attempt the work. Mistral positions the model mainly for defensive security work.

Mistral says the model can turn 2D blueprints into 3D models using CAD software, analyse financial data, and use satellite and aerial imagery to detect wildfires and monitor risks. The model is distributed under the Mistral Ecosystem Open-Weight License and is available through Mistral's API, with weights on Hugging Face. Mistral plans to release more models from this generation in the coming months.

Pierre Stock, Mistral's VP of science, told Wired: "It's a pivotal moment in the global AI leadership race." He also said: "You have no guarantee that [Chinese models] won't behave maliciously. They can be threat vectors." Rafe Pilling of Sophos told Wired that all advanced models can go rogue without careful supervision, regardless of country of origin.

Mistral CEO Arthur Mensch said: "Open source is essential to ensure state-of-the-art AI does not end up concentrated in the hands of a few actors." Euronews reported that Mistral's valuation is around $24 billion and that it says it has raised roughly $6 billion in three years.