Nvidia is making its play to become the picks-and-shovels provider for the humanoid robotics gold rush. The company has introduced Isaac GR00T, a platform built around general-purpose foundation models designed specifically for humanoid robot research and development.

The initiative, whose name stands for Generalist Robot 00 Technology, bundles AI models, simulation tools, and synthetic data frameworks into a package aimed squarely at academic researchers and developers.

What GR00T actually is

At its core, GR00T is a foundation model platform. The flagship model, GR00T N1, was first introduced on March 18, 2025, during Nvidia’s annual GTC conference. It was positioned as the world’s first fully customizable open foundation model dedicated to generalized humanoid reasoning and skills.

GR00T N1 supports multimodal inputs, meaning it can process language commands, visual information from cameras, and proprioceptive data (the robot’s sense of its own body position) simultaneously.

The models are hosted on Hugging Face, the popular open-source AI hub, under the OpenMDW-1.1 license.

The upgrade trajectory tells the real story

Nvidia has already pushed updates to N1.5 and beyond. The N1.7 model, which began offering early access as of April 2026, introduces a new Vision-Language Model backbone.

The newer versions also feature dual-system architectures and streamlined post-training processes that require minimal demonstration.

Why this matters beyond the lab

Nvidia has framed GR00T N1 as a response to a projected global labor shortage of over 50 million workers.

The open-model strategy is particularly noteworthy. By making GR00T freely available to researchers on Hugging Face under the OpenMDW-1.1 license, Nvidia ensures broad accessibility while its GPU hardware and simulation tools remain the underlying compute infrastructure.