BoxBot: Opening boxes with visual servoing and a VLA

BoxBot cuts tape with image-based visual servoing and opens flaps with an open-source VLA, fine-tuned on 125 teleoperated demonstrations collected in about five hours; no custom architecture and one data path built on Viam. Contact our team to learn more.

By submitting this form, you agree that Viam may use the information provided to respond to your request and contact you about Viam products and services. For more information, see our Privacy Policy. You may opt out of marketing communications at any time.

BoxBot Dataset on Hugging Face VLA Inference Service on GitHub Viam 101: Getting Started

Explore Viam

BoxBot Dataset on Hugging Face

The fine-tuned SmolVLA model and the 125-demonstration dataset behind BoxBot, open for training, benchmarking, or reproducing the setup.

Learn more

VLA Inference Service on GitHub

The service that runs BoxBot's fine-tuned VLA policy locally on the Jetson Orin Nano, open source and ready to deploy on your own setup.

Learn more

Viam 101: Getting Started

Build a palletizing cell in simulation with our free, 90-min self-paced course. Come away knowing the core of every Viam machine.

Start now