Introducing OM-1, our first robot foundation model, zero-shot generalizing to any robot: table-top arms, industrial arms and humanoids. - learned directly from human manipulation data - no teleop/robot data - close to human-level dexterity and efficiency - multi-robot collab

OM-1 captures subconscious human physical intelligence directly from humans: nudging, sliding, twisting, power grasping, and whole-body coordination. We build on our prior work DexCap ( at Stanford.

OM-1 zero-shot generalizes across different kinds of embodiments: small arms, industrial arms, and even humanoids, with our 7-DoF Omnibody Hands. No robot-specific finetuning needed. In a time when hardware advances every month, robot models and data need to be future-proof.

We built Omnibody Hand around function: useful contact points, in-hand reorientation, and smooth transitions between precision and power grasps. Its compact 7-DoF design combines thumb-index dexterity with coordinated finger motion for power grasps.

Efficiency is ROI. ROI is what gets robots deployed. OM-1 makes long-horizon tasks (quadmanual phone packaging, bartending, laundry folding) look like easy <30s jobs. All our videos are real-time with no speedup.

We also tested OM-1 on a deceptively hard task: unplugging an Ethernet cable. Unlike USB, Ethernet connectors lock firmly in place and release only when the latch is pressed very precisely. OM-1 reliably solves this fine-grained manipulation task.

OM-1 can power humanoids to do whole-body manipulation and navigation. More on this later.

OM-1 exhibits several emergent behaviors. Each arm learns to compensate for the other’s mistakes. The model knows when to retry, when to adapt to adversarial perturbations, and when to stop given the environment changes too drastically.

Please check out our website and blog for more information:

Congrats @zipengfu @chenwang_j !! So smooth!

Congrats! Super impressive

Super impressive.

Super impressive! Congratulations!

This robot is handling that cocktail shaker like it’s done this a thousand times.

Congrats @chenwang_j @zipengfu !!

💎💎💎

Wow - poor bartender. I would have loved to have a drink made by him and the robot to do the cleanup.

Humans make the dexterous task look unfairly easy. And that's what OM-1 and Omnibody Hand aims to capture.

Congrats @RewardAI_ ! No teleop, no on-robot data, and still human-speed execution across arms and humanoids — that's the claim I'd have bet against😍. Keeping each modality at its native sampling rate rather than resampling to a common clock seems like a big part of why it works at speed. Curious how the <30 min number holds across task families👀

Congrats @zipengfu @chenwang_j!

Gin &amp; Tonic ASAP plz! Congrats!

Can I get access to it? I want to try

当年的斯坦福做菜机器人，今天发布第一款大模型了

yawn. We've been demo'd to death. What we all want to hear: Price? Release date?

this looks interesting, congrats! would love to see more videos showing how actual data capturing looked like

Impressive hands

amazing speed, congrats on the launch!

Congratulations @zipengfu and @chenwang_j !

This is very impressive

Really impressive!

zero-shot across arms is a bold target. learning from human manipulation data makes the setup especially interesting.

Put the weights on huggingface

The zero-shot claim is the bit I’m watching. If OM-1 really carries the same manipulation policy from tabletop arms to humanoids, that’s a much bigger jump than another robot demo.

Really impressive work! Congrats @chenwang_j @zipengfu and team!

impressive and amazing！

So cute

This work is intresting because it directly focuses on cross embodiment problem which is cool.

@ChongZzZhang Is this available to try on open source arms?

I am extremely surprised if it’s seriously going at 1x

Super cool! Congrats 🚀🚀🚀 @zipengfu @chenwang_j

This is very interesting, I like!!

lovely

So smooth