Unitree R1-A5 vs R1-A7: Arms-First Home Robots

The reason is simple: most useful household jobs are not walking problems first. They are manipulation problems. A robot has to find a cup, reach around a chair, orient a wrist, close a gripper without crushing the object, recover when it slips, and do all of that safely near humans and pets. Legs help with stairs and whole-home coverage, but arms decide whether the robot can do anything once it arrives.

That is why Unitree's lower-cost dual-arm platform deserves a buyer-facing read. The R1-A5, R1-A7, R1-A5-D, and R1-A7-D are not home appliances. They are developer and research hardware. But they point at a path that may matter more than another bipedal demo: build useful arms first, then choose the right base.

What did Unitree actually announce?

Unitree announced the dual-arm R1 family on April 30, 2026 through its official social channel, describing it as a "Dual-Arm Humanoid Robot" starting from $4,290, with high-performance motors and sensors, binocular vision, voice interaction, and compact hardware for smarter task development. Unitree's official R1-D page now organizes the line into four configurations:

The shared claims are the interesting part. The family uses 5-DOF or 7-DOF arms, optional grippers or dexterous hands, binocular vision with RGB/depth interfaces, a 4-array microphone, dual speakers, 8-core body and head processors, 10 TOPS head-module compute, and optional NVIDIA Jetson Orin compute in higher configurations. Unitree's official page now lists 2-4 kg maximum arm payload depending on arm posture and a claimed ±0.1 mm end-clamp accuracy, though buyers should treat those as hardware claims to verify in task context, not proof that the robot can fold laundry.

Two caveats matter. First, Unitree says the line starts at $4,290, but the exact price for every configuration is not disclosed. The ui44 database entry for R1-A7-D therefore keeps price as unknown with a note that configuration pricing may require a quotation. Second, this is not a consumer-ready helper. Unitree's official R1-D page says the robot has a complex, powerful structure, asks users to maintain a sufficient safety distance from people, and notes that parameters may vary by configuration and scenario.

What Unitree R1-A5 vs R1-A7 specs matter for home buyers?

The headline difference is arm degrees of freedom. R1-A5 uses 5-DOF arms; R1-A7 uses 7-DOF arms. In plain language, more arm DOF usually means more ways to place the wrist and end-effector while keeping the elbow, shoulder, and robot body out of trouble.

That matters in homes because real homes are not clean factory cells. A robot may need to reach behind a mug handle, approach a drawer pull from the side, pick up a sock from a cluttered floor, or place an object on a shelf without bumping a lamp. A 5-DOF arm can still be useful for simple reach-and-place work. A 7-DOF arm gives the software more options when the scene is awkward.

The second difference is base choice. The non-D models are fixed-base platforms. They make sense for a workbench, a lab station, a kitchen-counter mockup, or any setup where the goal is to train arms without solving navigation. The D models add the wheeled chassis and chassis LiDAR. All four configurations are listed with external power or Li-ion battery support and about 1.5 hours of battery-powered runtime, so the D suffix is about mobility, not simply battery access.

That is the useful buyer translation:

• R1-A5: cheapest path to dual-arm experiments if the robot can stay in one place.
• R1-A7: better if the research question is wrist orientation, grasp pose, and dexterous end-effectors.
• R1-A5-D: mobile base plus lower arm complexity; good for navigation and simple manipulation data.
• R1-A7-D: the strongest signal for future home chores because it combines mobile reach with the more capable arm geometry.

None of those are a normal household purchase today. But if you are tracking which hardware will teach the industry how home chores actually work, R1-A7-D is the one to watch.

Why arms may matter before legs

Humanoid marketing tends to reward locomotion because it looks good on video. Walking up stairs, recovering from a shove, or doing a cartwheel is obvious and shareable. Home usefulness is quieter. It is the robot repeatedly picking up the same dropped toy without wedging its gripper under the sofa.

The Unitree R1 biped illustrates the split. In the ui44 database, R1 starts at $4,900 for R1 Air, with the standard R1 at $5,900. It stands about 123 cm, weighs about 29 kg, offers 20-26 DOF depending on configuration, and gets about 1 hour of mixed activity battery life. It is movement-first: walking, running, push recovery, cartwheels, handstands, voice/image interaction, OTA updates, ROS 2 support, and optional dexterous hands on EDU configurations.

That is exciting hardware. It is also not the same as a chore robot. Unitree's own official R1 page frames movement as the foundation for tasks, not proof that tasks are solved. The dual-arm R1 family flips the emphasis. Instead of asking, "Can this robot walk like a small humanoid?" it asks, "Can this robot reach, grip, perceive, and generate manipulation data cheaply enough to iterate?"

For buyers, this changes the questions worth asking. Do not start with whether a robot looks like a person. Start with whether it can safely perform the first inch of useful work: seeing the object, approaching it, orienting the hand, closing the gripper, lifting within payload limits, and recovering when the object moves.

How R1-A7-D compares with other manipulation-first home robots

R1-A7-D sits in an interesting middle ground. It is much cheaper than established research mobile manipulators, more manipulation-focused than Unitree's bipedal R1, and far less consumer-proven than a robot vacuum with a small arm.

Here is the ui44-style comparison:

The contrast with Stretch 3 is especially useful. Hello Robot Stretch 3 was designed around real homes from the start: 24.5 kg, 33 × 34 cm footprint, 141 cm height, 2 kg payload, 2-5 hour runtime, ROS 2 and Python support, web/gamepad/dexterous teleoperation, and a large indoor mobile manipulation research community. It is not a humanoid, but it has the shape of a practical home manipulator.

R1-A7-D is less proven and less home-tailored, but it is strategically important because it pushes dual-arm hardware into a much lower entry-price band. Even if only labs and developers buy it, the data they collect can shape the next wave of home robots.

The buyer trap: cheap arms are not the same as useful autonomy

The most dangerous reading of R1-A5/R1-A7 is "a $4,290 home robot is here." That is not what the evidence says.

The $4,290 figure is a starting price for the dual-arm line, not confirmed R1-A7-D retail pricing. The mobile models are heavier and more complex, while end-effectors, battery setup, and higher-compute Jetson Orin modules can vary by configuration. Warranty, shipping, service, and software support may differ by market. Most important, a manipulator platform does not automatically include the home autonomy people imagine when they hear "robot butler."

A useful home robot needs at least five layers to work together:

1. Hardware reach: Can it reach floor, table, counter, shelf, and handle heights without tipping or pinching?
2. End-effector fit: Is the gripper right for fabric, cups, packaging, and cluttered objects, or only demo parts?
3. Perception: Can it identify the object and its graspable area under bad lighting and partial occlusion?
4. Control: Can it move slowly, safely, and repeatably near people?
5. Recovery: When it drops the object, pushes it away, or gets blocked, does it ask for help, retry intelligently, or fail safely?

R1-A7-D is interesting because it can expose those layers to developers at lower cost. It is not interesting because it magically solves them.

What this means for home robot buyers in 2026

If you are shopping for a robot today, R1-A5/R1-A7 should probably not be on your shortlist unless you are a developer, robotics lab, university, integrator, or very technical early adopter. A normal household should look at proven products for specific jobs: a cleaning robot like Saros Z70 if the task is obstacle pickup plus vacuuming, or a mature mobile manipulator like Stretch 3 if the goal is research or assistive experimentation with support.

But if you are trying to understand where home robots are going, Unitree's move matters. It suggests the market is splitting into three practical tracks:

• Movement-first humanoids such as Unitree R1 and G1: exciting full-body platforms, useful for locomotion and whole-body AI work, but not automatically chore-ready.
• Manipulator-first platforms such as R1-A7-D and Stretch 3: less glamorous, but closer to the core problem of useful household work.
• Task-specific consumer robots such as Roborock Saros Z70: narrow, shippable products where the arm is constrained to one workflow.

The likely winner for homes may borrow from all three. A future home robot may need the mobility of a humanoid, the manipulation discipline of Stretch-style research platforms, and the consumer reliability expectations of a robot vacuum. Until then, arms-first robots are the better signal to watch because they attack the hard part directly.

Bottom line

Unitree R1-A5 and R1-A7 are not proof that affordable home humanoids have arrived. They are proof that affordable manipulation hardware is getting more serious.

For buyers, the R1-A7-D is the most relevant configuration because it combines a wheeled base, chassis LiDAR, 7-DOF arms, mobile perception, voice interaction, optional end-effectors, an official 2-4 kg arm-payload range that varies by posture, and an officially listed 1.5-hour battery-powered runtime. That is not enough for unsupervised household chores. It is enough to make developers ask better questions than "can it walk?"

If you want to compare the current evidence yourself, start with the ui44 pages for Unitree R1-A7-D, Unitree R1, Unitree G1, Hello Robot Stretch 3, and Roborock Saros Z70, or use the compare tool. The home robot story is not only about legs. It is about hands that can finally do useful work.