Physical Intelligence unveils OM-1 robot foundation model with zero-shot generalization
yifengzhu_ut · x · 2026-09-15
Physical Intelligence (RewardAI) introduced OM-1, its first robot foundation model that zero-shot generalizes to table-top arms, industrial arms and humanoids.
- Learned directly from human manipulation data — no teleop or robot data required
- Claims near human-level dexterity and efficiency
- Supports multi-robot collaboration
Industry observers (including Tongzhou Mu) called it a strong new entrant to the handheld gripper market. If the claims hold, training a general manipulation model without robot data is a notable challenge to standard robot-learning pipelines.
Related event: RewardAI Unveils OM-1, a Robot Foundation Model Trained Only on Human Data(40 posts)→
More from Embodied
- Robot neck design talk: pan motor in base, tilt in head for compact motion — carlosdponx · 2026-09-15
- NVIDIA Shows How to Train Healthcare Robots with GPU-Native Medical Physics Simulation — NVIDIA Developer · 2026-09-15
- OM-1 robot reliably unplugs locking Ethernet cables—a deceptively hard manipulation task — yifengzhu_ut · 2026-09-15
- 1,200 actuated cylinders form a programmable landscape for a robot in MuJoCo — hbouammar · 2026-09-15
- Arcturus launches 5K HDR color passthrough camera for Steam Frame, shipping this week — kscottz · 2026-09-15
- DIY pneumatic actuator expands 10.7mm from a 2.7mm resting height — IanPritchard · 2026-09-15