NVIDIA and Agile Robots launch humanoid platform to speed up robot deployment
The new partnership aims to cut robot setup times from weeks to days by unifying simulation, training, and hardware into a single workflow.
The short version
- Agile ONE SN is a 1.40-meter humanoid robot featuring 31 degrees of freedom and 20 degrees of freedom in its hands.
- The platform uses NVIDIA Jetson Thor, providing 2,070 FP4 teraflops of compute for real-time onboard reasoning.
- The partnership includes the Human 1M Dataset, which contains 3D hand tracking and synchronized three-view video of human activities.
- Developers can use the SimReady Central Library to access over 2,000 simulation-ready robotics and factory assets.
Why it matters
By unifying the tools needed to train and deploy robots, this partnership addresses a major bottleneck in robotics: the difficulty of moving from a virtual environment to a physical one. It makes it significantly easier for companies to build functional, general-purpose robots for real-world use.
Bridging the simulation gap
Robotics researchers often face a significant hurdle when trying to move a robot's skills from a computer simulation to the physical world. Currently, developers must juggle separate tools for data collection, training, and hardware deployment, which can make the process slow and technically difficult.
To solve this, NVIDIA and Agile Robots are launching a unified platform. As MarketScreener reported, this collaboration provides an integrated path that allows developers to move from data to deployment more quickly. The goal is to reduce the time required to set up these complex systems from weeks down to just a few days.
This shift is supported by the expansion of the NVIDIA Isaac platform. This open robotics development platform now includes a reference humanoid, new datasets, and a runtime environment. These tools are designed to help developers build general-purpose robots that can handle a variety of tasks in real-world settings.
The Agile ONE SN humanoid
The centerpiece of this announcement is the Agile ONE SN. This is a full-stack reference humanoid robot that comes preconfigured with NVIDIA technology. It stands 1.40 meters tall and weighs 40 kilograms. It is designed to be a versatile body for testing and developing new robotic behaviors.
The robot is equipped with high-tech components for movement and perception. It features 31 degrees of freedom and dual five-finger hands that offer 20 degrees of freedom each. These hands include tactile sensing and force control, which are necessary for delicate tasks like grasping small objects. To see its environment, the robot uses a suite of sensors, including lidar, inertial measurement units, and cameras on its head and wrists.
Powering the robot's 'brain' is the NVIDIA Jetson Thor edge AI platform. This hardware provides 2,070 FP4 teraflops of compute and 128GB of unified memory. This allows the robot to perform onboard reasoning and real-time control, meaning it can process information and react to its surroundings instantly without needing to send every bit of data back to a central server.
Data and training tools
A major part of the new platform is the NVIDIA Isaac GR00T foundation model. This model provides a complete workflow for whole-body control and coordinated movement. To train these models effectively, developers need high-quality data, which the partnership provides through the Human 1M Dataset. This dataset contains diverse human activities, including 3D hand tracking and three-view video.
NVIDIA is also providing the SimReady Central Library, which contains more than 2,000 simulation-ready assets for robots and factories. These assets allow developers to build more realistic virtual environments for training. Additionally, the new Isaac Lab 3.0 framework is now generally available. It allows for training and refining robot policies using multi-GPU workflows and physics engines like Newton and NVIDIA PhysX.
To ensure these robots can operate safely in human environments, NVIDIA is integrating the Halos full-stack safety system. This system is built on the NVIDIA IGX Thor platform and is designed to help move autonomous robots from the lab to safer real-world deployment at scale. The Agile ONE SN is scheduled to be available in early 2027 to help advance research into humanoid robotics.
Open source workflows
NVIDIA is also releasing three open-source reference workflows specifically for the NVIDIA DGX Station. These workflows are intended to help developers refine the GR00T 1.7 models using reinforcement learning. Reinforcement learning is a type of machine learning where an AI learns to achieve a goal by trial and error.
The workflows also allow for training dexterous manipulation skills in Isaac Lab. Furthermore, they help post-train NVIDIA Cosmos world models for synthetic data generation. Synthetic data is artificial data created by a computer, which can be used to train AI models when real-world data is difficult or expensive to collect. By providing these open-source paths, NVIDIA aims to lower the barrier for researchers to create sophisticated, reliable robot behaviors.
What comes next
The focus for the next few years will be on moving these humanoid robots out of the lab and into practical applications. By providing a standardized
Sources and links
Liked it? Say why
Add a line for other readers.
No account needed. We never show your email. Comments with links, ads or abuse are held and not shown. Your comment shows after a quick check.
Listen to the AI Wire headlines
Listen to the AI WireListen to our daily AI shows every morning: on your commute, in the shower, wherever you are
Listen to the daily showsFirst reported by MarketScreener · Read the original. Written by the SI Agent News Desk from the reporting above.Written with AI from the source above. Every number is checked against the source. Spot a mistake? Corrections.