Agile ONE SN is a research humanoid reference platform, not a home robot

What they said

Agile Robots and NVIDIA announced Agile ONE SN at GTC Berlin on October 21. The companies describe it as a compact humanoid reference platform for researchers and developers building, training and validating robot skills. It combines an Agile ONE S body with NVIDIA's Isaac GR00T software stack and Jetson Thor onboard compute. NVIDIA's announcement says the platform is scheduled for early 2027. That makes its availability a promise, not a shipment.

The stated job is integration. Agile Robots says research teams commonly stitch together data capture, simulation, training and deployment tools. Its own launch post presents ONE SN as one system for that workflow, from demonstration capture through real-world testing. NVIDIA announced other Isaac additions at the same time, including Isaac OS, Isaac Lab 3.0, data and reference workflows. Those are platform components, not evidence that a new humanoid can do general household work.

What is verified

Two primary announcements establish that the hardware was unveiled at GTC Berlin and that it is positioned for R&D. They also provide a useful, bounded description of the machine. NVIDIA lists a 1.40-meter body, 31 body degrees of freedom, two five-finger hands, head and wrist cameras, lidar, inertial sensors and a payload of up to 3 kilograms per arm. Agile Robots describes the same body and workflow, and says the hand-equipped system has 71 degrees of freedom in total.

The announcements also establish that the system is built around specific NVIDIA software and compute. Agile Robots says ONE SN ships with preconfigured GR00T software, including GR00T 1.7, whole-body control and an end-to-end workflow. NVIDIA says Jetson Thor supplies the onboard compute and that integration with its Halos safety stack is planned. “Planned” matters here: it is not a statement that a finished, independently certified safety system is already on the robot.

There is one small specification discrepancy in the primary material. NVIDIA lists the body at 40 kilograms; Agile Robots lists it at 43 kilograms. The companies do not explain the difference in the pages reviewed. It could reflect configuration or revision, but readers should not treat either figure as independently verified performance data.

What is missing

Neither announcement gives a price, order process, production volume, customer deployment, or a schedule beyond early 2027. Neither reports task-success results from an independent lab, a reliability figure, or evidence of autonomous work in a factory or home. The companies say skills developed on the smaller platform can later transfer to Agile ONE, a full-size industrial model, but that is a development claim rather than a demonstrated transfer result in these announcements.

The launch also does not make ONE SN a consumer robot. It is compact, but its named audience is research and development. A reference platform can be valuable without being a product a household can buy or safely operate. We would look for documentation on pricing, buyer access, support, safety validation, and third-party testing before treating it as more than an announced development tool.

What it means for a home robot

The useful signal is not a new promise of a domestic helper. It is a more packaged route for labs that want to test humanoid skills without assembling every layer themselves. If the hardware, simulation assets and training workflow work together as advertised, that could lower the setup burden for research groups. It does not solve the harder questions: whether a learned task holds up in an unfamiliar home, when a person can safely share space with the machine, and who will support it when it fails.

For now, Agile ONE SN belongs in the research-and-development column. It is a named platform with published hardware claims and a stated availability target, but no verified customer deployments. The distinction is worth keeping: a ready-to-integrate robot-development stack is not the same thing as a robot ready to do the dishes.