NVIDIA/Isaac-GR00T

NVIDIA Isaac GR00T N1.7 - A Foundation Model for Generalist Robots.

What it solves

NVIDIA Isaac GR00T N1.7 is an open vision-language-action (VLA) model designed to provide generalized humanoid robot skills. It enables robots to perform manipulation tasks across diverse environments by processing multimodal inputs—such as language and images—and translating them into continuous robot actions.

How it works

The model uses a combination of a vision-language foundation model (Cosmos-Reason2-2B / Qwen3-VL) as a backbone and a diffusion transformer (DiT) head that denoises continuous actions. A key innovation in N1.7 is the use of a relative end-effector (EEF) action space, which represents actions as deltas from the current pose rather than absolute targets. This allows the model to transfer manipulation priors learned from 20,000 hours of human video data (EgoScale) to robot control across different robot embodiments.

Who it’s for

This project is for robotics researchers and developers who want to deploy generalized manipulation skills on humanoid or semi-humanoid robots, as well as those looking to fine-tune VLA models on custom robot data.

Highlights

  • Cross-Embodiment Generalization: Uses a shared relative action space to work across different robot types.
  • Human-to-Robot Transfer: Pretrained on massive human video datasets to improve language-following and generalization.
  • Flexible Deployment: Supports zero-shot inference, fine-tuning with custom data, and acceleration via TensorRT.
  • Broad Integration: Compatible with Hugging Face LeRobot and supports various benchmarks like RoboCasa and SimplerEnv.

Related

  • Project
    dexmal/opendm

    OpenDM is a Vision-Language-Action (VLA) model framework for open-world robot control, providing tools for training, fine-tuning, and deploying models that can handle complex, multi-embodiment tasks.

  • Project
    Robbyant/lingbot-vla-v2

    LingBot-VLA 2.0 is a Vision-Language-Action foundation model that enables robots to generalize across different embodiments and tasks using a unified action representation and MoE action experts.

  • Dispatch
    Qwen-VLA: Unifying Vision-Language-Action Modeling for Embodied Intelligence

    Qwen-VLA is a general-purpose Vision-Language-Action model that unifies robotic manipulation, vision-language navigation, and cross-embodiment control into a single generalist policy model.

  • Project
    Open-X-Humanoid/HEX

    A whole-body vision-language-action framework for humanoid robots that enables cross-embodiment manipulation by aligning heterogeneous robot states into shared body-part slots.

  • Project
    2toinf/X-VLA

    X-VLA is a soft-prompted Transformer-based Vision-Language-Action model designed for scalable cross-embodiment robotic control across diverse hardware platforms.