GPT-6 Astra implements a quadruped locomotion policy directly, stretching the 'LLM' label

YuXiang_IRVL · x · 2026-09-09

After Prof. Malik questioned whether an LLM could directly implement a locomotion policy, the author tested it with GPT-6 Astra — and it worked on a quadruped robot. He argues 'LLM' no longer fits Astra, which is multimodal, reasoning, agentic, and spatially aware.

Related event: GPT-6 Astra Directly Controls Quadruped Locomotion(2 posts)→

Original post →

More from Embodied

Embodied channel →