A Guth Labs publication ↗
Back to News

ASUS puts Gallop on a factory task, but its performance bar is still a target

ASUS puts Gallop on a factory task, but its performance bar is still a target

ASUS has described its first humanoid robot, Gallop, learning to pack motherboard accessories inside an ASUS group electronics factory. The November 13 company account puts a specific industrial task ahead of a general-purpose robot promise: handling 15 small parts, including items in flexible transparent bags, at an existing workstation. ASUS sets a target of 99.5% success within 20 seconds. It does not report that Gallop has achieved it. ASUS, November 13.

The company says the current machine has two arms on a fixed base and runs on one onboard NVIDIA Jetson AGX Thor. Its training combines worker demonstrations, simulated variations and operator-guided practice. ASUS describes separate planning, supervisory, skills and control layers, allowing models to change without rebuilding the whole machine. These are company-reported design details, not an independently reproduced assessment. ASUS.

Customer pilots remain planned for 2027, with market availability targeted for the second half of 2028. Interesting Engineering independently reports those plans and the current factory test, but its account is based on ASUS rather than an outside measurement of reliability. Interesting Engineering, November 13, 03:35 EST.

Guth News analysis

The useful question here is not whether a humanoid can move a component once. It is whether the whole station can operate predictably without constant intervention. A narrowly defined packing job is a better basis for evaluation than a montage of unrelated demos, but the distinction between a target and a measured result is decisive.

For buyers, the next meaningful evidence would be sustained task success, intervention frequency, recovery from failed grasps, and independent customer experience. Those measures would help distinguish a promising engineering program from production-ready automation. The November 13 announcement establishes a program and a deployment ambition, not an available robot product or demonstrated replacement for a full shift of work.

Evidence limits and revision record

First coverage in this run. Actual trial start date is not specified; the program began in early 2026. November 13 is the dated public account, not a claimed first day of deployment. No pricing, independently verified success rate or external customer deployment was established. Earlier coverage was searched for Gallop and no matching accessible draft was found.

Sources and citations

Revision history

Published .