Mistral Large 4: Le Chonk Arrives
Mistral AI has launched a public preview of Mistral Large 4, unofficially nicknamed "le Chonk." It is a 1 trillion-parameter mixture-of-experts model with 49 billion active parameters, natively multimodal, and trained from scratch on 3'800 NVIDIA Grace Blackwell GPUs in Mistral's own European datacenters. Weights drop at the end of this month [1].
The model is already competitive with the strongest open-source models globally, and Mistral claims it significantly outperforms any open-weight model developed in the US or Europe. On critical enterprise workloads, including cybersecurity, finance, and law, they call it state-of-the-art among open models [1].
Cybersecurity: The Big Flex
This is where ML4 gets interesting. On the Artificial Analysis Cyber Index, an independent evaluation of how well AI models find and fix security flaws in real software, ML4 ranks among the top five models globally and leads open-weight models developed outside China by a wide margin [1].
On one test that asks the model to reproduce a real vulnerability in open-source software and then patch it, ML4 scores 82%, the highest of any model. Claude Opus 5.5 and GPT-6 Astra score near zero on the same test because they refuse to perform the task [1].
This is the core argument for open weights in security work. Defending software often starts with proving a flaw is real, exactly the kind of work safety filters in closed models block. As threat actors increasingly jailbreak closed models for offensive use, defenders need systems that can match those capabilities without being constrained by provider-level refusals. ML4 also solves 93% of challenges in Cybench, a set of 40 exercises drawn from security competitions [1].
Coding and Agentic Workflows
ML4 scores 49.8% on the combined Coding Agent Index, ahead of DeepSeek V4 Pro 0813 and Qwen3.8 Max. On DeepSWE v1.1 it hits 61.7%, on SWE-Atlas-QnA 59.4%, and on Terminal-Bench 4 it reaches 28.3% [1].
A blind human evaluation by Surge AI had professional annotators rate model outputs on a 1 to 5 scale with identities hidden. ML4 Preview ranked second of five at 3.74, ahead of Kimi K3 (3.59), GLM-5.3 (3.60), and GLM-5.2 (3.40), behind only Claude Opus 5 at 4.22 [1].
On AutomationBench, which covers 657 business workflows across Gmail, Google Sheets, Slack, and Salesforce, ML4 scores 59.9%, ahead of Kimi K3, MiMo-V2.6-Pro, and DeepSeek V4 Pro [1].
Multimodal
ML4 is a step change in visual understanding for Mistral's models. It reasons across complex documents, charts, and natural images. The model combines visual grounding with agentic capabilities, from inspecting gigapixel satellite imagery for disaster response to analyzing engineering drawings. Mistral claims it surpasses even frontier closed models in some visual grounding tasks [1].
Sovereignty and Infrastructure
The sovereignty angle is the real thesis. ML4 was trained on 3'800 Grace Blackwell GPUs in Mistral's own datacenters in Europe. The public preview is served on that same infrastructure. The model will be available across multiple regions worldwide, including a European deployment operated end-to-end by Mistral under European law [1].
Training data spanned more than 160 languages, including every official language of the European Union. Mistral recently raised €3 billion in a Series D round at a post-money valuation of more than €21 billion to push this strategy [1][2].
What to Watch
The open-weight release at the end of this month is the key event. Preview API benchmarks are one thing, third-party testing of the actual weights is another. If the cyber scores hold up under independent evaluation, ML4 becomes a serious tool for security teams that cannot afford a provider refusing mid-incident. The coding numbers are solid but not frontier-leading. The multimodal capabilities need external validation [1].
The model will also serve as the foundation for a new generation of specialized and optimized Mistral models, so the weights release is just the beginning [1].
Sources:
[1] Mistral AI - Introducing Mistral Large 4
[2] Mistral AI - Mistral raises €3B Series D at €21B valuation