Mistral Large 4 (le Chonk): Trillion-Parameter Multimodal Model

Mistral launched a public preview of Mistral Large 4 on 6 October 2026 (internally ML4, officially nicknamed le Chonk): Mistral says it is a 1-trillion-parameter natively multimodal model with about 52 billion active parameters, its largest and most capable model to date, available through a preview API on Mistral Studio the same day with weights dropping at the end of the month

Tool Overview

Features, steps and FAQ below

Features

✓ Frontier performance: Mistral says ML4 stands out across coding, agentic workflows and multimodal understanding, already competes with the strongest open-source models worldwide and significantly outperforms any open-weight model developed in the US or Europe, calling it state-of-the-art among open models on critical enterprise workloads including cybersecurity, finance and law
✓ Cybersecurity: Mistral says it ranks among the top five models globally on the Artificial Analysis Cyber Index, scores 82% on a test that asks a model to reproduce a real vulnerability and then patch it (the highest of any model) and solves 93% of Cybench; it says several leading closed models, including Claude Opus 5.5 and GPT-6 Astra, score near zero on the same test because they refuse to perform the task
✓ Agentic coding: Mistral reports 61.7% on DeepSWE v1.1, 59.4% on SWE-Atlas-QnA and 28.3% on Terminal-Bench 4, with a combined Coding Agent Index score of 49.8% ahead of DeepSeek V4 Pro 0813 and Qwen3.8 Max
✓ Multimodal and visual grounding: Mistral says it reasons powerfully across complex documents, charts and natural images and surpasses GPT-6 Astra on Dense 200 visual grounding (42% vs 41%), useful for inspecting gigapixel satellite imagery and verifying engineering drawings
✓ European build and sovereignty: Mistral says ML4 was trained from scratch on 3,800 NVIDIA Grace Blackwell GPUs in its own datacenters in Europe, with multilingual training data spanning more than 160 languages including every official EU language, and that a European deployment operated end-to-end by Mistral will be available

How to Use

Open Mistral Studio (console.mistral.ai) to see the Mistral Large 4 public preview
Call ML4 through the preview API and run a small evaluation on your security research or coding tasks first
Follow Mistral's blog for the end-of-month open-weight release and details on architecture, benchmarks and post-training methodology
Once weights are out, assess private cloud or on-premise deployment if you need self-hosting

FAQ

What is Mistral Large 4?

A public preview released by Mistral on 6 October 2026, described by Mistral as a 1-trillion-parameter natively multimodal model with about 52 billion active parameters, internally ML4 and nicknamed le Chonk.

Can I use it now?

Mistral says you can try the preview API today on Mistral Studio, with open weights dropping at the end of October 2026.

Will the weights be open?

Mistral says it will release the weights by the end of the month along with more details on the architecture, additional benchmarks and its post-training methodology, and that ML4 will serve as the base for a new generation of specialized Mistral models.

What makes it notable in cybersecurity?

Mistral says it ranks among the top five models globally on the Artificial Analysis Cyber Index and scores 82% on a reproduce-and-patch test, the highest of any model, and that several closed models score near zero there because they refuse the task.

How was it trained?

Mistral says ML4 was trained from scratch on 3,800 NVIDIA Grace Blackwell GPUs in its own European datacenters, with the preview served on the same infrastructure and training data spanning more than 160 languages.