On August 19, 2026, at the World Robot Conference (WRC) in Beijing Yizhuang, Galaxy General (Galbot) put three robot embodiments — the bipedal Galbot ET1, the mass-produced wheeled-arm Galbot G1, and the heavy-load Galbot S1 — on the same universal brain: AstraBrain-Agent. The S1 is not a demo: it is already running 7×24 hours in normal operation on a CATL production line.
ET1's Bipedal Debut: The Real Signal Isn't "It Can Stand"
After its Spring Festival Gala appearance, Galaxy General had been known for the wheeled-arm G1. ET1 marks the company's first engineering closure of "bipedal + same brain." The live demo was playing tennis — a task demanding whole-body coordination and real-time decision-making in adversarial conditions, far beyond pick-and-place.
- AstraBrain-Agent: handles task understanding, motion planning, and native physical-world interaction;
- Cerebellum modules: motion control, translating high-level intent into joint trajectories;
- "Sense, understand, decide, and execute concurrently": G1's capability in dynamic environments, already at industrial-grade commercial stability.
- 30 kg-class depalletizing, transport, and palletizing;
- 7×24 normal operation on a CATL production line;
- Partnerships with SAIC, BAIC, Bosch, Hyundai, and others.
- No need to pre-define responses for every disturbance;
- New objects and scenes don't require training a new specialized model;
- Deployers get a continuously adaptive general brain, not a single-station toy.
- ET1's "tennis" is a tech demo, not industrial-grade stability;
- "One brain, many bodies" has a sample size of three; whether it extends to robotic arms, quadrupeds, or cobots will be tested over the next 12 months;
- The 7×24 CATL figure lacks engineering metrics like continuous per-station hours or MTBF.
ET1 isn't trying to out-trick Optimus. The deliverable is one brain driving differently shaped bodies — an achievement few domestic companies have reached with bipedal, wheeled, and heavy-load robots simultaneously in service.
What "One Brain, Many Bodies" Actually Spans
AstraBrain overturns the industry's default fragmented approach of "one model per robot, one skill per scene":
| Embodiment | Commercial scenario | Current stage | |---|---|---| | Bipedal ET1 | Home service, tourism, reception | WRC debut 8/19, skills still being acquired (tennis etc.) | | Wheeled-arm G1 | Industrial manufacturing, retail | Mature and in mass production, used across customers | | Heavy-load S1 | 30 kg-class palletizing, logistics | 7×24 normal operation on CATL line |
Many international players have touted a "universal brain," but this is reportedly the first domestic case where the same model weights deliver productivity on three radically different embodiments.
S1 at CATL Runs the Numbers, Not a Demo
What S1 has already done:
The key detail for manufacturing: S1 at CATL is not a "trial" or "POC" — it is normal operation, meaning the line's process documents have incorporated the robot's cycle times, failure modes, and downtime policies. Otherwise CATL would not let it replace a fixed station worker.
What G1's Closed Loop Means for Engineers
G1 does not rely on pre-programmed responses to disturbances. Its single-model closed loop means:
This is engineering proof of embodied intelligence moving toward physical-world generality — the universal brain plus universal closed loop being the two capabilities that have blocked humanoid commercialization for a decade.
Reasons Not to Applaud Too Early
From "One Dish Per Robot" to "One Brain Cooking Every Cuisine"
The next industry question is no longer how many humanoids are built or how fast they walk, but how many robot forms a single set of model weights can cover. If within 12 months AstraBrain-Agent can drive 5+ embodiment types, win industrial orders beyond heavy loads, and let third-party robots plug in without retraining, "general embodied intelligence" will stop being a pitch-deck word and become infrastructure-grade — like SageMaker or Kubernetes. That bell has already rung on CATL's production line.