Milestones
- GO-1 (Genie Operator-1) unveiled with ViLLA frameworkMar 2025Complete.
- Deployed in A2 and X2Jun 2026Complete.
by AgiBotCN
Universal embodied foundation model using Vision-Language-Latent-Action architecture. Learns from human video and cross-embodiment data.
Foundation model deployed in A2 and X2 robots; enables rapid task learning.
Updated 2026Checked 10 Oct0 updates this week
Visual-language model processes robot perception and human instructions.
Mixture of experts for different motions; latent planner and action experts.
Real-world deployment data improves the model without full retraining.
| Spec | GO-1 (Genie Operator) |
|---|---|
| Training sources | Human video, robot dataR (reported) |
R reported by the company
AgiBot
A2 humanoid, Genie foundation model, GO-1
Chinese robotics company founded in 2023 by former Huawei engineers. The A2 is its flagship general-purpose humanoid; X2 is a compact research model. GO-1 is its universal embodied foundation model.