AGIBOT's GE-Act 2.0 claims first native World Action Model, success rate scales 17.1%→44.1% with 100x data
KrishRShah · x · 2026-09-12
AGIBOT released GE-Act 2.0, billed as the first native World Action Model to validate a pretraining and scaling path for embodied AI.
- Trained entirely from scratch on embodied manipulation data: visual representation, future generation, and action prediction — no inherited video generators, no task-specific fine-tuning.
- Data scaled 100x: from 300 to 30,000 hours.
- Tested zero-shot on real robots: unseen scenes and objects, 100 atomic tasks, 20 skill categories, two robot embodiments.
- G1-OP task success climbed from 17.1% to 44.1% with no sign of saturation, and new skills emerge at scale.
The result is a notable signal that pretraining + scaling can work for embodied manipulation.
More from Embodied
- OpenTac tactile sensing project to be showcased at CoRL — _krishna_murthy · 2026-10-28
- Hybrid LLM agent scores 83.2%, beats nine VLA policies on new SO-101 robot arm benchmark — Truman Hickok · 2026-10-28
- cixliv: robot fighting may be the first realized market where bots beat average humans — cixliv · 2026-10-28
- Can the Hiwonder MentorPi T1 run fully offline as a robot pet? Pi 5 questioned — oulu2006 · 2026-10-28
- Testing Mint Modeler to Build the Menlo Research Robot — freelerobot · 2026-10-28
- Khosla-backed startup builds a 'mind-reading' cap with 100,000 tiny chips to write AI prompts — Distinct-Question-16 · 2026-10-28