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
- J&J unveils Ottava surgical robot in push into lucrative new market — HealthcareLdr · 2026-09-12
- ESP32 Tip: Validate Features Locally via Copilot CLI Webpage Before Flashing the Device — lee_stott · 2026-09-12
- First hands-on calls the iPhone Duo the most promising first-generation foldable ever — BenBajarin · 2026-09-12
- Meta's slim Phoenix VR headset leaks ahead of Meta Connect — chrisgrayson · 2026-09-12
- Deployment evals seen as the key to closing robotics' demo-to-deployment gap — paigeinsf · 2026-09-12
- Analyst details Apple's S11, A20 and M6 silicon: new packaging, cooling and on-device AI designs — BenBajarin · 2026-09-12