The direct answer
Can integrated robot actuators be ranked by the largest torque number? No. Of the 18 selected records, 13 publish a continuous or rated value and five do not. Twelve records publish both a continuous-or-rated value and a peak value, allowing arithmetic inside that exact model. Four other records publish momentary stall torque while explicitly warning that it is not continuous output. One selected MYACTUATOR page exposes peak torque through the model code but no continuous-duty value in the cited text boundary.
What does the public evidence still omit? None of the 13 selected continuous-or-rated source rows states the rating duration, cooling method or exact temperature at which that torque was established. Eleven publish a separate operating ambient range, but an allowable ambient range is not the test temperature behind a torque rating. RCI therefore preserves the values as manufacturer claims and preserves the missing conditions beside them.
The twelve valid within-model peak-to-continuous-or-rated divisions run from 2.00× to 3.00×. These are arithmetic relationships between two labels in the same manufacturer table. They are not overload-duration guarantees, cross-brand scores or proof that either endpoint is simultaneously available with maximum speed.
Five labels describe five evidence boundaries
| Boundary | Records | Source family | What RCI can say |
|---|---|---|---|
| Continuous value published | 6 | HEBI Robotics | Six exact configurations publish separate continuous and peak rows. |
| Rated value published | 6 | CubeMars | Six exact configurations use “rated torque” and publish a separate peak row. |
| Manufacturer-calculated continuous value | 1 | ROBOTIS PH54 | The continuous value is calculated from core-motor specifications; the manufacturer asks long-duration users to consult it. |
| Stall only; no continuous value | 4 | Four ROBOTIS X-series records | The e-Manual calls stall torque momentary and distinguishes it from continuous and real-world output. |
| Peak only in cited text | 1 | MYACTUATOR X8-32 | Peak torque is encoded in the model definition; no continuous-duty value is exposed in the selected text boundary. |
“Continuous” and “rated” remain separate manufacturer labels. RCI does not rename CubeMars rated torque as continuous torque. The PH54 row is separate again because ROBOTIS says its continuous value is calculated from the core motor. This distinction matters when a buyer needs a thermally sustained joint output rather than a catalog field with a familiar name.
The 18-record evidence matrix
| Exact record | Duty boundary | Continuous / rated | Other headline torque | Within-model factor |
|---|---|---|---|---|
| HEBI Robotics R8-9+ R8-9+ column in the current R-Series+ family table | Continuous value published | 10 N·m Continuous torque | 23 N·m Peak torque · Peak duration not stated in the selected table | 2.30× |
| ROBOTIS XH540-W270-R XH540-W270-R at the 12.0 V catalog point; R suffix uses RS-485 | Stall only; no continuous value | Not published in the cited specification U-state boundary | 9.9 N·m Stall torque · 12.0 V, 4.9 A; momentary stall condition | — |
| CubeMars AK80-9 V3.0 KV100 AK80-9 V3.0 KV100 current product row | Rated value published | 9 N·m Rated torque | 22 N·m Peak torque · Peak current 28 A; duration not stated | 2.44× |
| MYACTUATOR RMD-X8-P9-32-R Current X8-32 V4 listing; P9 ratio and R interface encoded in the orderable model | Peak only in cited text | Not published in the cited specification U-state boundary | 32 N·m Peak torque encoded in model name · Manufacturer naming guide defines the final number as peak torque | — |
| HEBI Robotics T8-16+ T8-16+ column in the T-Series+ family table | Continuous value published | 20 N·m Continuous torque | 44 N·m Peak torque · Peak duration not stated in the selected table | 2.20× |
| HEBI Robotics H25-90 H25-90 column in the H-Series family table | Continuous value published | 90 N·m Continuous torque | 180 N·m Peak torque · Peak duration not stated in the selected table | 2.00× |
| ROBOTIS PH54-200-S500-R PH54-200-S500-R continuous row calculated by the manufacturer from core-motor specifications | Manufacturer-calculated continuous value | 44.7 N·m Continuous torque | 44.7 N·m Continuous torque · No separate peak or stall value selected for the current RCI comparison row | — |
| ROBOTIS XM540-W270-R XM540-W270-R at the 12.0 V catalog point; R suffix uses RS-485 | Stall only; no continuous value | Not published in the cited specification U-state boundary | 10.6 N·m Stall torque · 12.0 V, 4.4 A; momentary stall condition | — |
| CubeMars AK70-10 KV100 AK70-10 KV100 current product row | Rated value published | 8.3 N·m Rated torque | 24.8 N·m Peak torque · Peak current 23.2 A; duration not stated | 2.99× |
| CubeMars AK10-9 V3.0 KV60 AK10-9 V3.0 KV60 row in the current AK product comparison table | Rated value published | 18 N·m Rated torque | 53 N·m Peak torque · Peak current 31.9 A; duration not stated | 2.94× |
| HEBI Robotics T8-3+ T8-3+ column in the T-Series+ family table | Continuous value published | 4 N·m Continuous torque | 8 N·m Peak torque · Peak duration not stated in the selected table | 2.00× |
| HEBI Robotics T25-20 T25-20 column in the T-Series+ family table | Continuous value published | 20 N·m Continuous torque | 40 N·m Peak torque · Peak duration not stated in the selected table | 2.00× |
| HEBI Robotics H25-45 H25-45 column in the H-Series family table | Continuous value published | 45 N·m Continuous torque | 90 N·m Peak torque · Peak duration not stated in the selected table | 2.00× |
| ROBOTIS XH540-W150-R XH540-W150-R voltage-specific catalog points; R suffix uses RS-485 | Stall only; no continuous value | Not published in the cited specification U-state boundary | 6.6 / 7.1 / 8.5 N·m Stall torque · 11.1 / 12.0 / 14.8 V and 4.5 / 4.9 / 5.9 A respectively | — |
| ROBOTIS XW540-T260-R XW540-T260-R voltage-specific catalog points | Stall only; no continuous value | Not published in the cited specification U-state boundary | 8.8 / 9.5 / 11.2 N·m Stall torque · 11.1 / 12.0 / 14.8 V and 4.5 / 4.9 / 5.9 A respectively | — |
| CubeMars AK60-6 V3.0 KV80 AK60-6 V3.0 KV80 current product row | Rated value published | 3 N·m Rated torque | 9 N·m Peak torque · Peak duration not stated | 3.00× |
| CubeMars AK80-64 KV80 AK80-64 KV80 current product row | Rated value published | 48 N·m Rated torque | 120 N·m Peak torque · Peak current 19 A; duration not stated | 2.50× |
| CubeMars AKA60-6 KV80 AKA60-6 KV80 current hollow-shaft product row | Rated value published | 3 N·m Rated torque | 9 N·m Peak torque · Peak duration not stated | 3.00× |
The matrix is intentionally exact-model and source-bounded. HEBI family PDFs use model columns; CubeMars pages mix product-specific and family-comparison rows; ROBOTIS pages publish voltage-specific endpoints; the MYACTUATOR identity encodes ratio, peak torque and interface. A product-family label without its selected column, voltage or revision is not enough to reproduce the row.
A published ambient range is not a torque-rating temperature
For a thermal duty claim to be portable, a buyer needs more than a torque label. At minimum, the evidence should identify the duration or steady-state rule, cooling boundary, ambient and winding or case-temperature criterion, bus voltage, current limit, speed or duty cycle, controller derating and the exact hardware configuration.
| Condition checked across 13 published rows | Explicitly disclosed | RCI interpretation |
|---|---|---|
| Rating duration or steady-state qualification | 0 / 13 | The exact endurance or equilibrium rule is absent from the selected rows. |
| Cooling method or mounting heat sink | 0 / 13 | No selected row defines natural convection, forced cooling or a mounting thermal path. |
| Exact rating temperature | 0 / 13 | Operating ambient ranges are recorded separately and are not reclassified as rating-test temperature. |
| All three fields together | 0 / 13 | No selected public row supports a fully condition-matched continuous-output comparison. |
This does not prove that a manufacturer lacks internal qualification or that a published value is wrong. It means the selected public source does not let an outside reader reproduce the thermal boundary. The correct evidence state remains M, not V; the missing condition remains visible.
The 2.00×–3.00× spread is useful—inside each model
RCI calculated peak divided by continuous or rated torque only when both values appear for the same exact record:
| Record | Continuous / rated | Peak | Peak ÷ continuous/rated |
|---|---|---|---|
| HEBI Robotics H25-90 | 90 N·m | 180 N·m | 2.00× |
| HEBI Robotics T8-3+ | 4 N·m | 8 N·m | 2.00× |
| HEBI Robotics T25-20 | 20 N·m | 40 N·m | 2.00× |
| HEBI Robotics H25-45 | 45 N·m | 90 N·m | 2.00× |
| HEBI Robotics T8-16+ | 20 N·m | 44 N·m | 2.20× |
| HEBI Robotics R8-9+ | 10 N·m | 23 N·m | 2.30× |
| CubeMars AK80-9 V3.0 KV100 | 9 N·m | 22 N·m | 2.44× |
| CubeMars AK80-64 KV80 | 48 N·m | 120 N·m | 2.50× |
| CubeMars AK10-9 V3.0 KV60 | 18 N·m | 53 N·m | 2.94× |
| CubeMars AK70-10 KV100 | 8.3 N·m | 24.8 N·m | 2.99× |
| CubeMars AK60-6 V3.0 KV80 | 3 N·m | 9 N·m | 3.00× |
| CubeMars AKA60-6 KV80 | 3 N·m | 9 N·m | 3.00× |
HEBI’s selected ratios run from 2.00× to 2.30×; CubeMars’ selected ratios run from 2.44× to 3.00×. A larger ratio does not mean a stronger actuator, longer overload duration or better thermal design. Without a time-current-temperature envelope, it only says the two manufacturer labels are farther apart.
Five records correctly remain unknown for continuous output
- ROBOTIS XH540-W270-R: ROBOTIS explicitly says stall torque is momentary and differs from continuous and real-world output; no continuous-duty number is substituted.
- MYACTUATOR RMD-X8-P9-32-R: The selected page identifies peak torque in its model definition but does not expose a continuous-duty torque value in the cited machine-readable evidence boundary; image-only parameters are not silently transcribed.
- ROBOTIS XM540-W270-R: The e-Manual warns that stall torque differs from continuous and real-world output; the separate independent deployment record does not validate continuous torque.
- ROBOTIS XH540-W150-R: The three voltage-specific stall endpoints are momentary; the e-Manual does not publish continuous output torque in the selected specification.
- ROBOTIS XW540-T260-R: IP68 does not turn the published momentary stall endpoints into a continuous-duty output rating.
Catalog v0.4.3 adds explicit U-state continuous-torque claims to XH540-W270-R and MYACTUATOR X8-32, aligning them with the three existing U rows for XM540-W270-R, XH540-W150-R and XW540-T260-R. An empty comparison cell can look like an ingestion failure; an explicit unknown states that RCI checked the cited source and found no adequate value.
What to request before using a torque rating
- Exact assembly: motor winding, reducer ratio, encoder, brake, connector, firmware and controller-current configuration.
- Thermal boundary: ambient, mounting plate, heat sink, airflow, coolant, case and winding limits, and derating behavior.
- Time boundary: continuous steady-state criterion, peak duration, recovery time and permitted repetition.
- Electrical boundary: bus voltage, current limit, commutation and control mode, and whether values are DC, phase RMS or controller ADC units.
- Mechanical operating point: output speed, efficiency, direction, external load, radial or axial load and gearbox temperature.
- Evidence: a curve or log from the exact actuator assembly—not only the core motor—plus acceptance limits for production units.
For safety-relevant sizing, validate the expected robot trajectory and worst-case thermal environment on the exact hardware. A public catalog row is an input to that test plan, not its replacement.
Method and limitations
RCI selected all 18 records in the integrated-joint actuator comparison group, checked the cited current manufacturer page or PDF on 2026-08-10, preserved the manufacturer’s own torque labels, and marked whether the selected source disclosed duration, cooling method and exact rating temperature. Operating ambient ranges and electrical rows were retained separately so they could not be mistaken for the rating test.
The audit is a public-evidence review, not dynamometer testing and not a census of every actuator on the market. RCI calculated only within-record peak-to-continuous-or-rated ratios. It did not calculate mechanical power from peak torque and no-load speed, infer missing continuous values, or compare stall-only records with rated records.
Download the record-level data
Download the immutable research dataset as JSON or CSV. Stable current aliases are /data/research/actuator-torque-boundaries.json and /data/research/actuator-torque-boundaries.csv. The two explicit unknown additions are in catalog v0.4.3 JSON and its claim CSV.
Suggested citation: Robot Component Index, “Robot Actuator Torque Claims: 18 Records, 5 Duty Boundaries,” RCI 015, research dataset v0.1.0 and component catalog v0.4.3, 2026-08-10.