NVIDIA Unifies Four AI Model Families Under Linux Foundation's OpenMDW-1.1 License

News Summary

The Linux Foundation released OpenMDW-1.1 on May 28, 2026 (Eastern Time), and NVIDIA announced it will adopt this open licensing framework across future releases of its Cosmos, Isaac GR00T, Ising, and Nemotron AI model families — a move that signals a significant step toward standardized, permissive licensing for large-scale AI models.

What Is OpenMDW?

OpenMDW stands for Open Model, Data and Weights. It is a permissive open license specifically designed for machine-learning models and their associated artifacts, including model architecture, weights and parameters, source code or scripts, documentation, and training data components. Originally co-created by the Linux Foundation and the PyTorch Foundation in 2025, OpenMDW addressed a long-standing gap in the open-source ecosystem: traditional open-source licenses were written for software code and did not adequately cover the complex multi-component nature of modern AI models.

Unlike earlier attempts at AI licensing — such as various community licenses attached to large language models — OpenMDW contains no copyleft or share-alike conditions. This means derivative models and integrations built on top of OpenMDW-licensed materials can remain fully permissive, reducing legal friction for downstream developers, enterprises, and researchers.

OpenMDW-1.1: What Changed

The 1.1 update refines the language of the original specification to provide greater clarity around scope and permissiveness. OpenMDW-1.1 grants unrestricted, royalty-free permission to "deal in the Model Materials without restriction, including under all copyright, patent, database, and trade secret rights included or embodied therein." Crucially, the license does not mandate which components a model publisher must release — it governs only how others may use whichever components are published. This flexibility makes it suitable for both fully open releases and partial-artifact distributions.

The license was formally published on May 28, 2026 (Eastern Time) and is now available for any AI model provider or developer to adopt worldwide.

NVIDIA's Adoption: Four Model Families Covered

NVIDIA's commitment to OpenMDW spans four distinct open model families, each targeting a different frontier of AI research and deployment:

• Cosmos — A world-foundation model designed for physical AI and simulation, enabling robots and autonomous systems to reason about and interact with the physical world.
• Isaac GR00T — A humanoid robot foundation model that underpins NVIDIA's robotics platform, allowing researchers and engineers to develop general-purpose robotic behaviors.
• Ising — NVIDIA's quantum-computing-inspired model family, focused on combinatorial optimization problems relevant to logistics, drug discovery, and financial modeling.
• Nemotron — A family of large language models optimized for enterprise agentic workflows, offering strong reasoning and instruction-following capabilities.

By adopting OpenMDW-1.1, NVIDIA standardizes the legal framework governing all future releases of these families under a single, model-centric license. Developers who use these models gain explicit rights to train, fine-tune, modify, redistribute, and deploy them — rights that were previously scattered across bespoke, per-model license documents.

Why Standardized AI Licensing Matters

The fragmentation of AI model licenses has been a practical obstacle for enterprise adoption and collaborative research. Organizations integrating multiple open models into a single pipeline had to track and comply with a patchwork of different terms, attribution requirements, and use restrictions. OpenMDW consolidates this into one clear framework, lowering the legal overhead for building on open AI foundations.

The Linux Foundation's involvement provides governance credibility. As the steward of projects such as the Linux kernel, Kubernetes, and the PyTorch Foundation, the Linux Foundation brings a well-established track record of maintaining neutral, community-governed open standards. Hosting OpenMDW under this umbrella ensures that the license specification itself can evolve through open community processes rather than unilateral vendor decisions.

Industry Implications

NVIDIA's adoption is expected to catalyze broader ecosystem uptake. Model hubs, cloud providers, and enterprise AI platforms that currently host NVIDIA open models may align their own distribution terms with OpenMDW-1.1, potentially creating a more uniform licensing landscape across the AI model supply chain.

For developers and researchers, the practical impact is straightforward: less time reading legal documents, more certainty about what they are permitted to do with model weights, and a clearer path to building commercial products on top of open foundations. The explicit inclusion of patent rights — unusual in many open-source licenses — is particularly valuable for organizations concerned about IP exposure in AI deployments.

Broader Open-Source AI Context

The OpenMDW release coincides with a period of intensifying discussion around what "open" truly means for AI models. Debates over whether releasing weights alone constitutes openness — versus also releasing training data and code — have prompted calls for more granular and honest licensing frameworks. OpenMDW responds to this by scoping itself precisely: it governs the artifacts a publisher chooses to share, without dictating what must be shared, and it does so with language that practitioners and legal teams can interpret consistently.

With NVIDIA's endorsement and the Linux Foundation's governance, OpenMDW-1.1 is poised to become a reference standard for responsible, permissive open AI licensing across the industry.