Figure AI Unveils Helix 02: The Next-Generation AI Brain Enabling Full-Body Humanoid Control
January 28, 2026 — Figure AI has officially released Helix 02, its next-generation vision-language-action (VLA) AI system, marking a significant leap in autonomous humanoid robotics. The new system extends from upper-body control to full-body coordination, fundamentally transforming how humanoid robots perceive, think, and act.
Revolutionary Architecture: Helix 02 introduces a groundbreaking three-layer neural network architecture:
| System | Role | Speed | Parameters | |--------|------|-------|------------| | System 2 | Scene understanding, language | 7-9 Hz | 7B | | System 1 | Full-body joint targets | 200 Hz | 80M | | System 0 | Whole-body balance & coordination | 1,000 Hz | 10M |
System 0 is trained on 1,000+ hours of human motion data and replaced 109,504 lines of hand-engineered C++ code with a single neural network.
Key Capabilities:
- Zero-Shot Generalization: Handles thousands of unseen household objects with natural language commands
- Full Autonomy: 4-minute continuous dishwasher task with zero human intervention, 61 separate actions
- Multi-Robot Collaboration: Multiple Figure 03 robots work together like human teams
- Edge Deployment: Complete model runs on low-power embedded GPU, no cloud dependency
Real-World Validation: Figure 02 robots have been deployed at BMW's Spartanburg plant since January 2024, contributing to the production of 30,000+ vehicles with >99% placement accuracy.
