#

Key takeaways

• Mistral Large 4 is in public preview.

• It is a 1 trillion-parameter natively multimodal model.

• The model has 49 billion active parameters.

• It targets coding, agentic workflows, and multimodal understanding.

• Weights are planned for release by the end of the month.

Mistral Large 4 arrives in public preview

Mistral has launched a public preview of Mistral Large 4, also called ML4. The company says users can try the preview API today on Mistral Studio. It also says the weights will drop at the end of the month.

ML4 is described as the company's largest and most capable model to date. It is a 1 trillion-parameter natively multimodal model with 49 billion active parameters. Mistral says the model continues to improve as it is refined.

What the model is designed to do

The source says ML4 shows strong performance across coding, agentic workflows, and multimodal understanding. It also says the model is competitive with the strongest open-source models globally. Mistral further states that it significantly outperforms any open-weight model developed in the US or Europe.

The company also says ML4 is state-of-the-art among open models on critical enterprise workloads. These workloads include cybersecurity, finance, and law. In some areas, such as visual grounding, Mistral says the model goes further and surpasses frontier closed models.

Release plan and current access

Mistral says the weights will be released by the end of the month. Until then, the company is red-teaming the model in real-world settings. It says this work involves cybersecurity leaders, vetted partners, and state authorities.

Those groups will access the same model with reduced moderation and expanded cyber capabilities. The source presents this as part of the testing process before the wider release. No further details are provided about the scope of those tests.

Training and infrastructure

ML4 was trained from scratch on 3,800 NVIDIA Grace Blackwell GPUs. Mistral says the training took place in its own datacenters in Europe. The public preview is served on the same infrastructure.

The company frames ML4 as a milestone in its long-term investment across infrastructure, research, and product development. It says the goal is to deliver state-of-the-art performance in critical verticals through open weights. It also says the model is designed to give customers control over their AI.

Why this launch matters

This announcement combines three themes: scale, openness, and enterprise focus. Mistral is positioning ML4 as a large multimodal model with strong performance in practical workloads. It is also emphasizing open weights and customer control.

The source does not provide independent benchmarks or technical documentation beyond these claims. Readers should treat the announcement as the company's own description of the model's capabilities. The most concrete next step is the public preview API, followed by the planned weight release.

Morocco relevance

The source reports no Morocco-specific fact. For readers in Morocco, the global lesson is to watch how open-weight models are being positioned for enterprise and multimodal use.

What to watch next

The main follow-up is the weight release at the end of the month. Another point to watch is how the model performs once more users test it outside the preview. The source also suggests that red-teaming and controlled access will continue before broader availability.