Delta Intelligence

Forging General Intelligence for Humanoids

[Delta Intelligence @delta_intelli]

Full story: https://t.co/obaumBWyJo

If what we're cooking excites you, please let us know
📮 [email protected]

[Siyuan Huang @siyuanhuang95]

Super excited to introduce Δ₀ (Delta-0), Delta Intelligence’s first humanoid foundation model (HFM)! 🤖

Δ₀ tackles some of humanoid robotics’ toughest challenges: autonomous, long-horizon execution across a room, dexterous manipulation across many degrees of freedom, and coordinated whole-body movement.

One robot puts on a record and dances to the music. Another gets to work on the household chores. 🎶

We want humanoids that can dance, work, and share moments of joy with us—becoming part of everyday life.

Above all, we hope this video sparks a little wonder and gets you excited about a future with robots alongside us. ✨

Check out the blog at https://t.co/toSaSw1c1G. We hope to bring humanoid intelligence into the next chapter, where we can scale the data, evaluation, and experiences for humanoid robots toward more capable HFMs.

[Delta Intelligence @delta_intelli]

Full story: https://t.co/obaumBWyJo

If what we're cooking excites you, please let us know
📮 [email protected]

[Delta Intelligence @delta_intelli]

Introducing Δ₀ (Delta-0), our humanoid foundation model for whole-body loco-manipulation.

One generalist policy. 69 degrees of freedom. Everyday tasks at near-human speed—from making a bed to putting on a record.

Watch Δ₀ in action.

[Delta Intelligence @delta_intelli]

Real-world RL learns from autonomous attempts and human-in-the-loop corrections.

In a mirrored kitchen, dishwasher success rose from 20% (4/20) to 65% (13/20) after RL post-training with targeted human guidance.

[Delta Intelligence @delta_intelli]

Learning from people, at scale.

Δ₀ is pretrained on 10,000+ hours of paired egocentric observations and whole-body motion, with a shared action space across embodiments.

Meet Delta D1, our panoramic full-body data capture device:
https://t.co/BEB0r6GbL6

[Delta Intelligence @delta_intelli]

Δ₀ pairs a latent world-action model with a learned 69-DoF controller.

The brain models visual futures and actions. Delta-action inputs let people correct motion during execution while the controller maintains whole-body coordination.

[Delta Intelligence @delta_intelli]

What can a humanoid do in your home?

Long-horizon tasks. Fully autonomous. One uncut take.

Our first full demo. Coming soon.