Mistral Large 4 (Le Chonk) Open Weights Release

Overview

Mistral AI has officially released the weights for Mistral Large 4, internally codenamed "Le Chonk." This 1-trillion-parameter, natively multimodal model is designed to serve as a foundation for sovereign AI, with a specific focus on coding, cybersecurity, and reasoning efficiency. The release includes a shift in pricing and a commitment to providing organizations with the ability to self-host and customize their AI infrastructure.

Coding and Reasoning Performance

Mistral Large 4 has seen significant performance gains since its public preview, driven by three weeks of additional reinforcement learning. The model now leads several open-weight benchmarks, including FrontierCode 2.0 and Terminal-Bench 4.0. Notably, the model provides high-level coding performance while maintaining token efficiency; it achieved a 48% score on the Artificial Analysis Coding Agent Index while using 47% fewer tokens than comparable models. To manage the balance between performance and inference costs, the model now supports three distinct reasoning modes: Low, Medium, and High.

Cybersecurity Capabilities

Designed with defensive security in mind, Mistral Large 4 is optimized for vulnerability discovery, reproduction, and patching. It achieved a 51% score on the Artificial Analysis Cyber Index. The model includes built-in guardrails that distinguish between defensive and malicious requests, maintaining a 96.6% refusal rate for harmful cyber prompts. Mistral is also offering a specialized variant of the model to enterprise partners for advanced security workflows, such as red-teaming and penetration testing, under strict contractual safeguards.

Sovereignty and Deployment

A core pillar of the Mistral Large 4 release is the promotion of sovereign AI. By providing open weights, Mistral enables organizations to inspect, customize, and self-host the model on their own infrastructure, reducing dependency on external APIs. For those requiring managed services, the model is available via Microsoft Foundry and other cloud providers. The model is released under the "Mistral Ecosystem Open-Weight license," a modified MIT license that permits free use and modification, with a requirement for a separate commercial license for entities operating a Model-as-a-Service (MaaS) business.

Community and Industry Reception

The release has sparked significant discussion regarding the definition of "open" versus "open-weight" models. Critics point out that the requirement for a separate commercial license for MaaS providers prevents the model from being classified as true open-source by OSI standards.

"Calling this 'open weight' is misleading. That's why we had huge backlash when Redis, Elasticsearch, and Terraform went source-available, right? Weight-available is a better adjective for this kind of licensing."

Despite licensing debates, many users view the model as a crucial alternative for organizations with strict data residency or provenance requirements, particularly those unable to utilize models developed outside of Europe or the US.

"Almost every regulated business in the US cannot run Chinese models, which was the only real option for self-hosted inference and agents... a near frontier-level model like this being available opens a lot of doors."

Developers have also noted that the pricing structure is significantly more competitive than the initial preview, with current rates set at $0.70 per million input tokens and $2.10 per million output tokens.

Sources

Related

  • Dispatch
    Mistral Large 4 (Le Chonk) release notes

    Mistral AI has released Mistral Large 4, a 1-trillion-parameter open-weight multimodal model optimized for coding, cybersecurity, and reasoning efficiency.

  • Dispatch
    Mistral Large 4 Release Notes

    Mistral Large 4 (ML4), a 1 trillion-parameter multimodal model with 49 billion active parameters, is now in public preview with open weights scheduled for release by the end of October 2026.

  • Dispatch
    Mistral Large 4 Release Notes

    Mistral AI has launched a public preview of Mistral Large 4, a 1 trillion-parameter natively multimodal open-weight model designed for frontier performance in cybersecurity, agentic coding, and professional knowledge work.

  • Dispatch
    Mistral Large 4 Release Notes

    Mistral AI has released Mistral Large 4, an open-weight multimodal model featuring a 1.05T parameter Mixture-of-Experts architecture optimized for cybersecurity and visual grounding.

  • Dispatch
    Mistral Large 4 model announcement and community reaction

    Mistral AI released Mistral Large 4, a 52‑billion‑parameter open‑weight multimodal model with a 1.6 B vision encoder, prompting mixed reactions about its performance, cost, and data policies.