Figure Ai · 4× · tolest 04. Juli 2026
Helix is a proprietary Vision-Language-Action model (VLA) by robotics company Figure AI that combines perception, language understanding, and motion control for humanoid robots in a single neural network. Introduced in February 2025, it has since controlled the upper body and, more recently (with Helix 02), full-body control of Figure robots (Figure 02/03), including grasping unfamiliar objects via voice command and parallel control of multiple robots. The model runs entirely on-board on embedded, energy-efficient GPUs and is primarily deployed in logistics, warehouse, and household applications. Helix is not a separately marketed product with public pricing, but is tightly integrated with Figure's own robot hardware and commercial customer partnerships.
Features
| Deployment-Modell | Onboard-Ausführung auf robotereigener Hardware (kein Cloud-Zwang), proprietäres Modell exklusiv für Figure-Roboter |
| Einsatzbereich | Haushaltsaufgaben (Einräumen von Lebensmitteln, Geschirrspüler be-/entladen), Logistik/Lager (Paketsortierung), industrielle Fertigung |
| Integrationen | Integriert in Figure-Roboter Figure 02 und Figure 03; nutzt Kopfkameras, Handflächenkameras und taktile Fingerspitzensensoren als Eingaben |
| Plattform | Läuft onboard auf eingebetteten, stromsparenden GPUs in Figure-Humanoidrobotern (Figure 02/03) |
| Release-Datum | Erstveröffentlichung 20. Februar 2025; Helix 02 am 27. Januar 2026 |