GPT-6 Astra as a Quadruped Locomotion Policy: 250 Inferences, 5 Seconds of Walking
YuXiang_IRVL · x · 2026-09-09
Responding to a professor's challenge that an LLM could directly implement a quadruped locomotion policy, Srinivas ran the experiment with GPT-6 Astra: the model output joint targets at 50 Hz like an RL policy, controlling a simulated Unitree Go1 with physics paused between calls. 250 inferences produced 5 seconds of walking — a playful test of whether frontier models, if orders of magnitude faster and cheaper, could serve as edge-device control policies.
Related event: GPT-6 Astra Directly Controls Quadruped Locomotion(2 posts)→
More from Embodied
- CosmoH2G: dataset and baseline for transferring hand demos to robot grippers — Hongxiang Zhao · 2026-09-09
- TANGO: whole-body VLA model navigates humanoid robots in cluttered spaces from sim data — Anqi Li · 2026-09-09
- whurley rides CyberCab robotaxi again: 'clean and comfortable,' calls it his next car — whurley · 2026-09-09
- Apple's foldable iPhone named iPhone Duo, $2,000 starting price, October launch — petefang · 2026-09-09
- Hanshow and X-EraLab bring embodied retail robots to nearly 70,000 stores worldwide — 量子位 · 2026-09-09
- Hyper3D WorldGen turns a single photo into a fully editable 3D scene — dr_cintas · 2026-09-09