DigestAI news desk

Cut through the AI noise.

Robotics & Physical AI1 min read

Stanford and Caltech use GPT-6 Astra in Unitree G1 robot to clean unfamiliar kitchen

Researchers from Stanford and Caltech introduced HomeBody, a system that lets a Unitree G1 robot explore an unknown kitchen, build a digital twin in Nvidia's Isaac Sim, and autonomously tidy up and fetch items. The robot relies on a swappable vision‑language model—GPT‑6 Astra—that calls directly into an extensible skill library for grasping, navigation, and drawer opening, eliminating the usual…

1 source

Key points

  • Stanford and Caltech built HomeBody, letting a Unitree G1 robot autonomously clean an unfamiliar kitchen.
  • The system connects GPT-6 Astra directly to the robot, bypassing a trained control layer.
  • Limitations noted include model latency, overheating finger servos, and high compute costs.

The language model plans each cleaning step and self‑corrects when errors occur. The team noted practical limits: GPT‑6 Astra’s response latency, overheating of the robot’s finger servos, and the high compute cost of running the model. The code has been released on GitHub, and OpenAI has previously signaled a return to robotics research for personal‑use applications.

Model page: GPT-6 (Astra) →

Full story from The Decoder · by Matthias BastianOpen source ↗

Researchers plug GPT-6 Astra directly into a robot and let it clean up an unfamiliar kitchen

The Decoder · 27 September 2026

Loading the full article…

This text was published by The Decoder and written by Matthias Bastian. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗

Topics · follow one to build your own front page

The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.

Comments

via GitHub Discussions

More in Robotics & Physical AI

All →

Related stories