A human hand has roughly 27 degrees of freedom, and closing that gap has become the real bottleneck determining whether a humanoid robot can actually work alongside people — Tesla's Optimus Gen 3 hands reach 22 degrees of freedom per hand through a biomimetic, tendon-driven design, while Unitree's G1 ships with two genuinely different hand options depending on how much dexterity you're willing to pay for. Raw degrees-of-freedom counts get thrown around as the go-to benchmark in this space, but they don't tell the whole story, and it's worth understanding both the numbers and their limits.
Tesla Optimus Gen 3: Doubling Down on Biomimetic Design
Tesla's Gen 3 hand system, revealed in detail through 2026, represents a genuine redesign rather than an incremental update. Each hand carries 22 degrees of freedom — double the 11 DoF of the Gen 2 hand — achieved through 25 actuators per forearm, for 50 total actuators across both hands. The key engineering decision was relocating every one of those actuators out of the hand itself and into the forearm, connected through a tendon-pull system that mimics how human hands work: muscles in the forearm pull tendons that move the fingers, rather than cramming motors directly into the fingers themselves. That relocation keeps the hand lightweight, compact, and cooler than a design with motors packed inside it would allow. Tesla also added tactile fingertip sensors with genuine force feedback, letting the hand sense and adjust grip pressure rather than operating purely on pre-programmed motion. Combined with more than 28 degrees of freedom in the body, Optimus Gen 3 reaches over 72 total degrees of freedom, with production-ready hands and factory deployment targeted for the second and third quarters of 2026.
Unitree G1: Two Very Different Hand Options
Unitree takes a genuinely bifurcated approach depending on configuration and budget. The base G1's standard three-finger Dex3-1 hands are force-controlled, meaning they can sense and adjust grip pressure in real time — but the company is direct about their real limits: adequate for grasping objects, but explicitly inadequate for fine manipulation tasks like threading, button pressing, or tool use requiring human-level precision. For genuine five-fingered dexterity, Unitree offers a separate EDU Ultimate configuration with dexterous hands contributing to a total of 43 degrees of freedom across the whole robot — but that capability comes at a real cost difference: the base G1 starts around $16,000, while the EDU Ultimate variant with true dexterous hands runs $73,900, more than four times the price specifically to unlock meaningfully better hand function.
Figure and the Rest of the Field: Converged on Five Fingers, Diverged on Everything Else
As of April 2026, the major humanoid developers — Tesla, Figure, Sanctuary AI, Boston Dynamics, and Apptronik among them — had broadly converged on five-fingered, anthropomorphic hand designs as the default approach, a meaningful consensus after decades of research hands taking different shapes. Where they diverge is actuation method, sensing technology, materials, and control architecture, and detailed public specifications aren't uniformly available the way Tesla has disclosed its own numbers — Figure's Helix-generation hands are described in the same anthropomorphic five-finger category, but without the same granular actuator count and tendon-routing detail Tesla has published. It's worth noting this isn't a brand-new problem: the research-grade Shadow Dexterous Hand, first released in 2005, already achieved 20 degrees of freedom with tendon-driven actuation and extensive tactile sensing — but cost more than $100,000 per unit and remained too fragile for real deployment, which is exactly the combination of dexterity, durability, and manufacturability today's humanoid companies are still racing to solve simultaneously.
The Honest Caveat: DoF Count Isn't the Whole Story
Degrees of freedom gets cited constantly as the primary benchmark for robot hand capability, but treating it as the full picture is a mistake worth avoiding. Higher DoF counts theoretically enable more complex movement, but they don't by themselves guarantee reliable real-world performance — a hand with fewer degrees of freedom but better force control, tactile feedback, and manufacturing consistency can outperform a higher-DoF hand that's less reliable in practice. Industry analysts covering this space have specifically flagged a cautious editorial stance toward DoF-based marketing claims, since the number alone says nothing about actual grip strength, durability under repeated use, or how well a hand performs on the specific tasks a factory or warehouse actually needs.
Robot Hand Specs at a Glance
| Robot | Hand configuration | Degrees of freedom | Notable feature |
|---|---|---|---|
| Tesla Optimus Gen 3 | Five-finger, tendon-driven, forearm-mounted actuators | 22 per hand (50 total actuators) | Tactile fingertip force feedback |
| Unitree G1 (base) | Three-finger Dex3-1, force-controlled | Not specified for fine manipulation | Adequate for grasping only, per company |
| Unitree G1 (EDU Ultimate) | Five-finger dexterous hands | 43 total across full robot | 4.5x the base model's price for the upgrade |
| Human hand (reference) | Five fingers, muscle-tendon actuation | ~27 | Baseline every humanoid hand is measured against |
Frequently Asked Questions
How many degrees of freedom does Tesla Optimus's hand have?
The Gen 3 hand has 22 degrees of freedom, double the 11 DoF of the Gen 2 hand, achieved through 25 actuators per forearm relocated out of the hand itself using a tendon-driven design, similar in principle to how human forearm muscles control finger movement.
Can Unitree G1's standard hands do fine manipulation tasks?
Not reliably, according to the company itself. The base three-finger Dex3-1 hands are described as adequate for grasping but inadequate for fine manipulation like threading or button pressing. Unitree's separate EDU Ultimate configuration adds true five-finger dexterous hands for a significantly higher price.
Is a higher degrees-of-freedom count always better for robot hands?
Not necessarily. Higher DoF theoretically enables more complex movement, but doesn't guarantee reliable real-world performance on its own. Force control, tactile feedback, and manufacturing consistency all matter alongside the raw DoF number, and industry analysts caution against treating DoF as a complete capability benchmark.
How does the human hand compare to current robot hands?
The human hand has roughly 27 degrees of freedom, along with thousands of tactile receptors and fine force control that current robot hands still don't fully match. Tesla's Optimus Gen 3, at 22 DoF per hand, is among the closest current designs to that biological benchmark.