Reward AI launches OM-1: one robot policy zero-shot across arms, industrial arms and humanoids
zipengfu · x · 2026-09-15
Reward AI (RT'd by founder Zipeng Fu) unveils its Omnibody stack and OM-1, built on the principle "One Model, One Data Interface, Any Body":
- Omnibody Hand: a wearable mocap device, building on Stanford's DexCap work (with Li Fei-Fei and C. Karen Liu), capturing dexterous manipulation at the pace people actually work
- One data interface: the hand serves as a shared learning interface across robot bodies
- OM-1: a general-purpose policy learned from human behavior, running zero-shot on small arms, industrial arms, and humanoids with 7-DoF Omnibody Hands—no robot-specific finetuning
The team argues human-level manipulation requires co-designing data, learning, and control rather than adding more data or compute, making models future-proof as robot hardware iterates monthly.
More from Embodied
- Fei-Fei Li's World Labs Raises $1.2B, Unveils 3D World Model Atlas — lukas_m_ziegler · 2026-09-15
- Hackathon Project Skillio Turns Vision Pro Into an AR Screw-Driving Coach Agent — seanmcdonaldxyz · 2026-09-15
- Autonomous robot unloads trucks without any fixed infrastructure — TinfoilTricorn · 2026-09-15
- Tesla FSD handles full descent on Slovenia's highest road at 2,055 m — FlorianGallwitz · 2026-09-15
- Physical Intelligence unveils OM-1, a robot foundation model trained on human data — kenziyuliu · 2026-09-15
- Unitree's new dual-axis robot neck mocked: tilt-then-pan kinematics force absurdly long neck — carlosdponx · 2026-09-15