← Leaderboard

Genie Envisioner (GE-Act)

World-action
AGIBOT · Aug 7, 2025

World foundation platform from AGIBOT: a video world model (GE-Base) pretrained on AgiBot World, a flow-matching action decoder (GE-Act) that turns its latents into robot actions, and a neural simulator (GE-Sim). Results are self-reported on AGIBOT’s own real-robot tasks, where the paper says GE-Act outperforms π0, UniVLA and GR00T N1 (figures only, no table). GE-Base and a CALVIN-trained GE-Act checkpoint are released.

Type
World-action
Parameters
—
Backbone
LTX-Video 2B (GE-Base) + 160M action decoder
Weights
Open
License
CC BY-NC-SA 4.0
Commercial use
Not allowed
Released
Aug 7, 2025
Robots
AgiBot G1, AgileX Cobot Magic (ALOHA), Dual Franka
SO-101
–
LeRobot
–
Inputs
RGB · P · L
VRAM
–

Benchmark results

0 of 13 benchmarks

No benchmark results recorded yet.

Not reported on 13 benchmarks
Wrong or missing number? Report an error