Twenty-two pieces put through the same four-protocol wringer, across four platforms buyers are shortlisting in mid-2026. One piece stood out for the wrong reason (it survived the abuse). The rest of the honours are inside. No sponsorships. No affiliate arrangements. No fee for coverage, ever.
Read this issue's Editor's Choice ⟶Best-in-test for guest-facing Optimus deployment. Cut against the Gen 3 shoulder-and-hip pattern; holds through a full hospitality shift; reads unambiguously as premium attire.
USD 6,800+ · atelier direct · four-month lead Value pick 02 / Figure 03The dressing that ships with the platform. Native to the shell architecture, engineered against the platform's own kinematic profile.
Bundled with platform · second colourway shipping Under $500 03 / Unitree G1Best-in-test for the sub-USD-500 category. Three interchangeable outer layers; printed head accessories; the version 3 update improved shoulder clearance meaningfully.
USD 320 · direct from maker Heavy-lift 04 / Atlas (electric)Designed against the electric Atlas's 50 kg lift capacity. Rated for extended outdoor use. Not a guest-facing piece; not intended to be one.
USD 2,400+ · contract onlyThis issue was made by
| Platform | Piece | Best for | Editors' rating | Estimated price |
|---|---|---|---|---|
| Tesla Optimus (Gen 3) | Paris-cut charcoal three-piece | Guest-facing hospitality, retail concierge | ★★★★★4.8/5 | USD 6,800+ |
| Figure 03 | Figure-native textile shell, Ivory | Enterprise + home general-purpose | ★★★★★4.6/5 | Bundled |
| Unitree G1 | Hong Kong hobbyist kit, v3 | Research, education, ambient demos | ★★★★☆4.4/5 | USD 320+ |
| Atlas (electric) | Engineered technical coverall, Cat. IV | Heavy industrial, back-of-house | ★★★★☆4.4/5 | USD 2,400+ |
The most common mistake we saw across the twenty-two pieces we tested this quarter was not a bad piece of tailoring; it was a good piece of tailoring cut for the wrong platform. Buyers procuring for a Figure 03 who bring in a piece cut for the Optimus Gen 3 will discover, roughly two hours into the first live shift, that the shell architecture is different in ways their atelier partner did not fully understand at the fitting stage. The reverse is also true: an Optimus Gen 3 in a Figure-native shell is a garment covering the wrong body. The correct sequence is always: platform first, atelier second, dressing brief third. If the atelier your property has a relationship with cannot demonstrate delivered work on the specific platform you are procuring, ask them to defer the commission until they can. This is not a hypothetical caution. Across the twenty-two-piece test bench this issue, three otherwise-good pieces were disqualified from consideration because the atelier had cut them against the wrong platform pattern block. The atelier in each case was a well-known name we would happily commission for a different platform. Platform fit is not fungible; a good couture house is not automatically a good couture house for your specific humanoid.
Nine pieces cut for the Optimus Gen 3 pattern, evaluated across four protocols. Editor's Choice, three runners-up, one value pick and a piece we recommend against.
The Figure 03 shipped with a design decision that transforms what "clothing for a humanoid" actually is. We evaluate the launch pieces and preview the third-party pipeline.
Modular fabric shells, printed accessories, aftermarket kits for the G1's compact form. Priced for research programmes, classroom deployments, and ambient hospitality demos.
Two ends of the market, one Optimus Gen 3, one working eight-hour shift at a partner property. The couture piece wins on merit; the mass piece wins on price-to-scale. What the specific tradeoff looks like in ratings, numbers, and photographs.
Atlas is a back-of-house platform first, and its clothing decisions reflect that. We evaluate the engineered technical coverall category and its emerging alternatives at working payload.
Static fit at three positional standards. Dynamic wear across an eight-hour service simulation. Blind guest reception at a partner property. Ten-cycle donning-and-doffing durability. Ratings weighted by intended use case. Weights published inside each review.
Read the full methodology →No placement fees. No commissions. No affiliate arrangements with any humanoid platform manufacturer, atelier, or contract producer. Corrections labelled at the head of the affected review. Institutional supporters disclosed annually.
Read the editorial policy →Twice-monthly, from Paris and Tokyo. Editorial calendar in advance, review notices, corrections, and the occasional editor's note. Unsubscribe with one click; no dark patterns.