Figure Helix: first generalist VLA driving a full-body humanoid
In one sentence Figure announces Helix, a proprietary Vision-Language-Action model controlling the Figure 02 humanoid at 200Hz, two robots in collaboration, fingers included. Demos: fold laundry and tidy a kitchen from language alone.
Figure, one of the most-watched humanoid startups, splits with OpenAI (which supplied their language models until 2024) and announces Helix: their proprietary "Vision-Language-Action" (VLA) model — a single network that takes images + voice commands and produces motion directly.
What's new: Helix doesn't just drive the arms but the whole body (35 degrees of freedom, fingers included) at 200Hz, and it can coordinate two different robots on the same task ("one hands me the apple, the other puts it in the basket").
In the demos, operators just say "put the groceries away" and the robot opens the fridge, recognizes vegetables it's never seen, decides where they go. It's the jump from "robot executing pre-recorded trajectories" to "robot that understands and improvises".
Companies
Figure
Tools
Helix VLA, Figure 02
Tags
Sources