The Physical Envelope: Size, Weight, Joints, Reach and Force
Physical specifications are the easiest numbers to publish and the easiest to misread, because almost all of them are quoted at their best single condition.
Every physical number on a humanoid datasheet describes a condition. A payload is a payload at a posture, for a duration, at a temperature. A degree-of- freedom count says how many joints exist, not which of them can be commanded independently or held under load. A reach figure rarely says what the robot can do at the edge of it.
Grouping these five topics makes the trade visible. Mass, reach, joint count and force all draw on the same actuators and the same battery, so a platform that leads on one usually concedes on another. The buyer's job is to find out which concession was made, and whether it lands on the task in hand.
How to use these comparisons
The market mixes several kinds of product. The Unitree G1 is a compact platform whose base and EDU editions differ in ways that matter. H1 and H2 are full-size machines with different configurations and handling demands. 1X NEO is presented as an Early Access home robot with ownership and subscription paths and a managed assistance model. Comparing them is only useful once the buyer has stated whether the goal is research, application development, industrial evaluation or household service.
For every topic below, write the requirement without adjectives such as advanced, safe, fast or production-ready. State an observable condition, a measurement, a tolerance and an evidence source, and treat a blank cell in a seller's response as a condition of purchase rather than an optimistic assumption.
- Dimensions and workspace
- Weight and recovery
- Degrees of freedom
- Payload and reach
- Torque and force
Dimensions and workspace
The question this topic has to settle is standing height, folded envelope, turning clearance, storage, doorways, floors and safe test-cell layout. That sounds narrow, but it crosses mechanics, software, service and contract boundaries: a published maximum may describe one option, one motion or a laboratory condition, while a buyer needs to know what the delivered robot repeats, under what limits and with what remedy when the answer differs from the quotation.
G1's smaller body changes transport and lab layout, H1/H2 require full-size handling and recovery planning, and NEO's domestic positioning still leaves a powered machine moving through human spaces. This is a procurement distinction, not a claim that one platform is better everywhere.
Minimum evidence packet. dimensioned drawings for the quoted revision, a measured route from loading dock to test area, floor loading review, exclusion-zone plan and recovery-equipment clearance. Ask the seller to mark each item standard, optional, unavailable or buyer-supplied.
How this goes wrong. the crate reaches the building but the robot cannot safely pass a doorway, turn in the test area, lie down for service or be recovered after a fault
Weight and recovery
The question this topic has to settle is mass, center of gravity, carrying, lifting, fall recovery and the human factors of moving an unpowered robot. That sounds narrow, but it crosses mechanics, software, service and contract boundaries: a published maximum may describe one option, one motion or a laboratory condition, while a buyer needs to know what the delivered robot repeats, under what limits and with what remedy when the answer differs from the quotation.
the roughly 35-kilogram G1 and the heavier full-size H1/H2 create different lifting hazards, while NEO's softer domestic design changes but does not eliminate manual-handling and blocked-joint scenarios. This is a procurement distinction, not a claim that one platform is better everywhere.
Minimum evidence packet. verified mass with options, lift points, recovery fixtures, two-person procedures, transport case details, stair and elevator limits and a disabled-robot drill. Ask the seller to mark each item standard, optional, unavailable or buyer-supplied.
How this goes wrong. operators improvise a lift after a fall, damaging joints or injuring themselves because the purchase included motion instructions but no safe recovery method
Degrees of freedom
The question this topic has to settle is joint counts, optional axes, usable ranges, coupling and why degrees of freedom do not equal task capability. That sounds narrow, but it crosses mechanics, software, service and contract boundaries: a published maximum may describe one option, one motion or a laboratory condition, while a buyer needs to know what the delivered robot repeats, under what limits and with what remedy when the answer differs from the quotation.
G1 configurations span materially different joint totals, H1 and H2 publish their own body architectures, and NEO's task usefulness depends on control quality and end effectors rather than a single joint-count comparison. This is a procurement distinction, not a claim that one platform is better everywhere.
Minimum evidence packet. the exact kinematic tree, joint limits, torque-speed curves, collision geometry, calibration process, API exposure and a task-level reachability test. Ask the seller to mark each item standard, optional, unavailable or buyer-supplied.
How this goes wrong. a procurement table ranks the highest joint count first even though critical wrists, hands or developer controls are optional or inaccessible
Payload and reach
The question this topic has to settle is rated payload, reach envelope, two-hand handling, balance margins and the difference between lifting and completing a task. That sounds narrow, but it crosses mechanics, software, service and contract boundaries: a published maximum may describe one option, one motion or a laboratory condition, while a buyer needs to know what the delivered robot repeats, under what limits and with what remedy when the answer differs from the quotation.
G1's compact scale, H1/H2's full-size geometry and NEO's household manipulation produce different useful envelopes that cannot be reduced to one arm-load figure. This is a procurement distinction, not a claim that one platform is better everywhere.
Minimum evidence packet. payload definition, wrist and arm limits, center-of-mass assumptions, reach map, grasp fixture, walking-with-load trial and repetitive-cycle temperature data. Ask the seller to mark each item standard, optional, unavailable or buyer-supplied.
How this goes wrong. a robot lifts an object once near its torso but cannot pick it from the required shelf, turn, walk and place it reliably
Torque and force
The question this topic has to settle is joint torque, endpoint force, continuous versus peak output, contact limits and how power affects risk as well as capability. That sounds narrow, but it crosses mechanics, software, service and contract boundaries: a published maximum may describe one option, one motion or a laboratory condition, while a buyer needs to know what the delivered robot repeats, under what limits and with what remedy when the answer differs from the quotation.
full-size Unitree platforms publish striking peak joint figures, G1 operates at a different scale, and NEO emphasizes low-inertia, compliant domestic interaction rather than industrial peak force. This is a procurement distinction, not a claim that one platform is better everywhere.
Minimum evidence packet. continuous curves, duty cycle, thermal limits, software caps, stop distance, measured endpoint forces and confirmation that the tested safety settings match production. Ask the seller to mark each item standard, optional, unavailable or buyer-supplied.
How this goes wrong. the team designs around a peak joint number that cannot be sustained, does not map to endpoint payload and increases the severity of contact
This guide uses manufacturer material checked on August 30, 2026: Unitree G1 product page, Unitree H1 product page, Unitree official store, 1X NEO product and ordering page, 1X legal terms. Prices, configurations, delivery queues, software access and specifications change. Product pages document what a manufacturer currently represents; they are not independent performance tests, and they are not a substitute for the final quotation and terms.
Bottom line
Treat every physical figure as a claim with conditions attached, and ask for the conditions in writing before treating it as a requirement.
The single most useful test is unglamorous: have the robot perform your actual motion, with your actual payload, at your actual duty cycle, and measure what happens on the fiftieth repetition rather than the first.