$1300 open source robot hands

What happened

A team released detailed plans and code for building a dexterous robot hand that costs less than $1,300 and can now move its wrist and spread its fingers like a human hand can. This means anyone with a 3D printer and basic tools can build a hand capable of learning complex manipulation tasks, instead of being locked into expensive proprietary hardware.

$1,300 cost to build complete hand

11 degrees of freedom in original Ruka

2 new degrees of freedom added in wrist

51.3% reduction in task completion time versus Ruka

21.2% increase in success rate versus Ruka

13 dexterous tasks demonstrated

assumed Dexterous robot hands required expensive, proprietary hardware or lacked the range of motion needed for human-like tasks.

found An open-source design with wrist mobility and finger abduction can be built for under $1,300, complete 13 dexterous tasks faster, and performs 51.3% quicker than the prior version.

Why this matters

For years, robot learning has been bottlenecked by hardware cost and access. A single commercial dexterous hand costs $10,000–$50,000, which means only well-funded labs can run experiments. Open-source hardware at $1,300 with full documentation changes the economics of who can iterate on robot control. The addition of wrist mobility and finger abduction matters because those are the degrees of freedom that separate 'can grasp a block' from 'can manipulate objects in tight spaces and rotate them in hand' — the difference between a toy and a tool. Watch whether academic labs and robotics startups actually adopt this design and whether it accelerates the pace of published results in dexterous manipulation.

The Shadow Hand costs over $100,000. Ruka-v2 costs $1,300. Both pick up thin objects. The difference is not the hand.

Who wins, who loses

who wins Robotics labs without capital budgets, who got a complete dexterous manipulation platform posted online, quietly, by an NYU team funded by Honda and Hyundai.

who loses Proprietary hardware vendors whose pricing model depended on researchers having no alternative at this capability level.

also Anyone running a robot learning lab on a grant budget, and the four competing labs in Switzerland, Germany, and the US who dropped nearly identical hands in the same twelve months.

Jargon decoder

dexterity ability to perform precise, complex hand movements

tendon-driven actuated by cables and pulleys that mimic how human tendons work

degrees of freedom independent directions the hand or fingers can move

abduction/adduction spreading fingers apart and bringing them together

The signal

Why this hasn't landed yet

It is a hardware paper with a GitHub link, which reads as a technical release rather than an event. Nothing failed visibly, no company lost money on a specific date, and the significance is cumulative — it only matters if you know what a Shadow Hand costs.

What happens next

Within the next twelve months, at least one of the competing open-source hands — ORCA, Aero, ISyHand — cuts its price or capability gap in response, and the community benchmark for dexterous manipulation tasks migrates away from proprietary hardware entirely.

The catch

Proprietary hardware vendors do nothing, because open-source lab hands rarely survive contact with industrial tolerances and support contracts, and the Allegro Hand remained widely adopted for a decade despite being, by the field's own description, unreliable and non-anthropomorphic.

The longer story

The thing the document buries

Ruka-v1 had exactly 11 degrees of freedom distributed as 2 per finger and 3 at the thumb, a very specific architectural choice that the new version must have modified.

The longer arc

The Yale OpenHand Project made the argument for open-source dexterous hardware around 2013. It took roughly twelve years, one price compression from $100,000 to $16,000 to $2,000 to $1,300, and four simultaneous competing releases for the argument to become a race.

Part of a pattern

Fourth open-source dexterous hand released in roughly twelve months, following LEAP Hand (CMU, 2023), ORCA Hand (ETH Zurich, April 2025), Aero Hand Open (TetherIA, late 2025), and ISyHand (Max Planck, September 2025). The compression is not slowing.