If you are buying an arm for a university lab, a robotics course or an imitation-learning setup, the price ladder runs from about $300 to about $50,000 — and unlike the industrial cobot market, it does not behave. Spend six times more and you can end up with half the payload. That is not a pricing error. It is the whole structure of this segment, and understanding it is the difference between buying a teaching toy for research work or a research platform for a teaching lab.
Three days ago I priced the industrial cobot payload ladder at a single dealer and found dollars-per-kilogram falling monotonically, from $12,304/kg on a 3 kg UR3e down to $1,987/kg on a 30 kg UR30. Bigger arm, cheaper capacity. Clean curve.
Run the same arithmetic across research arms and the curve inverts in the middle.
The ladder
| Arm | Price | DOF | Payload | Reach | Repeatability |
|---|---|---|---|---|---|
| Elephant myCobot 280 M5 | **$649** | 6 | 250 g | 280 mm | — |
| ROBOTIS OpenMANIPULATOR-X | **$1,629.09** | 4 + gripper | 500 g | 380 mm | **< 0.2 mm** |
| Elephant myCobot 320 M5 | **$2,279** | 6 | 1 kg | — | — |
| igus ReBeL (6-DoF) | **$7,499** | 6 | — | — | — |
| ROBOTIS OMY-3M | **$11,498.85** | — | — | — | — |
| Kinova Gen3 Lite | **~$13,500** with gripper | 6 | **0.5 kg** | 760 mm | — |
| ROBOTIS OpenMANIPULATOR-P | **$16,176.02** | — | — | — | — |
| Franka Research 3 (FR3) | **~€25,466 (≈$27,000–30,000)** | 7 | 3 kg | 855 mm | **< ±0.1 mm** |
| Kinova Gen3 (7-DOF) | **~$30,000–35,000** | 7 | ~4 kg | — | — |
*Prices for Elephant, ROBOTIS, igus, Kinova and Franka are the figures GrabaRobot verified on manufacturer and dealer pages between July 15 and July 31, 2026 (see our individual brand price guides). Specifications are from manufacturer documentation; where a manufacturer does not publish a figure, the cell is blank rather than estimated.*
Finding 1: dollars per kilogram is non-monotonic, and the peak is in the middle
- Elephant myCobot 320 M5 — $2,279/kg
- ROBOTIS OpenMANIPULATOR-X — $3,258/kg
- Franka Research 3 — ~$9,000/kg
- Kinova Gen3 Lite — ~$27,000/kg
The most expensive kilogram on this list belongs to a mid-priced arm. A Gen3 Lite costs roughly 5.9× a myCobot 320 and lifts half as much. If you shop this segment the way you would shop an industrial cobot — payload first, then price — you will conclude the Kinova is absurd, and you will be wrong, because payload is not what its buyers are paying for.
Finding 2: repeatability doesn't explain the gap either
The obvious next hypothesis is precision. It survives about one line of arithmetic.
ROBOTIS documents the OpenMANIPULATOR-X at < 0.2 mm repeatability, 4 DOF plus gripper, 500 g payload, 380 mm reach, built on DYNAMIXEL XM430-W350-T actuators — at $1,629.09. Franka documents the Research 3 at < ±0.1 mm, 7 DOF, 3 kg, 855 mm — at roughly $27,000.
Twice the positional precision for about sixteen times the money. Repeatability is a real difference and it matters for machine-tending style tasks, but it is nowhere near a 16× difference. Something else is being sold.
Finding 3: above roughly $10,000, you are buying the control interface
Here is what actually separates the tiers, and it is barely visible on a spec sheet.
Franka publishes that the FR3 has seven integrated torque sensors — one at every joint. That is the line item. Joint-level torque sensing is what makes impedance control, contact-rich manipulation, force-controlled assembly and safe physical human-robot interaction possible at all. An arm without it can be *positioned* precisely; it cannot be made *compliant* in a way you can model and publish about. The seventh degree of freedom does related work: redundancy lets you hold an end-effector pose while moving the elbow, which is the entire point for motion-planning research.
Kinova's Gen3 Lite sits at $13,500 with 0.5 kg of payload for the same category of reason — it is a supported research platform with a documented low-level API, not a payload device. The 5.4 kg arm weight and 760 mm reach tell you what it is designed around: mounting on mobile bases and wheelchairs, where every gram of arm mass is a cost and 500 g of grip is genuinely enough.
Meanwhile the sub-$2,500 tier is honest about what it is. A myCobot 280 at $649 gives six axes, a real SDK and 250 g of payload — an excellent way to teach kinematics, ROS basics and pick-and-place logic to thirty students at once. It will not give you publishable force control.
Finding 4: the actuators are most of the cheap arms
There is a useful sanity check available in this segment that you cannot run on an industrial cobot: ROBOTIS sells the servos its own arms are built from. Five XM430-W350-T actuators — the OpenMANIPULATOR-X's joints — cost $1,551.95 at the company's own catalogue prices, against $1,629.09 for the complete arm. The frame, gripper and assembly account for $77.14, under 5% of the price.
That is the structural reason budget research arms cluster where they do. Below about $2,500, you are essentially buying a bag of smart servos with a bracket set. The value added by the manufacturer is small, which is why the prices are competitive and why the capability ceiling is hard.
How to choose
- Undergraduate teaching, 10+ units, kinematics and ROS fundamentals → myCobot 280 M5 at $649, or OpenMANIPULATOR-X at $1,629.09 if you want the tighter repeatability and the ROBOTIS/TurtleBot ecosystem. Buy quantity, not capability.
- Graduate coursework and light research, one shared arm → myCobot 320 ($2,279) or igus ReBeL ($7,499). Real payload, real workspace, no torque sensing.
- Mobile manipulation, assistive robotics, arm-on-a-base → Kinova Gen3 Lite. You are paying for low mass and a supported API; do not benchmark it on payload.
- Force control, contact-rich manipulation, anything you intend to publish → Franka Research 3 or Kinova Gen3 7-DOF. The joint torque sensors are the purchase; everything else is packaging.
- Imitation-learning data collection → budget for *pairs* of arms, not one. This is the one application where the sub-$2,500 tier scales better than the premium tier, because you need leader and follower arms and lots of them.
Browse the wider category on our robot arm page.
The mistake I see most often is a lab buying one $27,000 arm when its actual research programme needed four $2,300 arms, or buying six $649 arms and then discovering that the paper it wants to write requires torque sensing. Decide which of those two failure modes would hurt more, and buy against that.
Sources
- Franka Robotics product specifications (FR3: 3 kg, 855 mm, < ±0.1 mm, 7 torque sensors)
- ROBOTIS e-Manual, OpenMANIPULATOR-X specifications
- RobotShop, Kinova Gen3 Lite 6-DOF arm, 0.5 kg payload
- GrabaRobot, ROBOTIS price guide (July 30, 2026)
- GrabaRobot, Elephant Robotics price guide (July 29, 2026)
- GrabaRobot, Kinova robot arm price guide (July 18, 2026)
- GrabaRobot, Franka Robotics price guide (July 16, 2026)
- GrabaRobot, igus ReBeL price guide (July 17, 2026)
- GrabaRobot, cobot payload class price curve (July 28, 2026)



