Finder
What robot vacuum should I buy? The home-fit finder
Verdict
Swept's home-fit finder checks 24 exact US robot vacuums against your own home, and its headline finding is that a bigger suction claim doesn't mean more suction.
- MOVA V70 Ultra Complete 42,000 0.6
- Roborock Saros 20 36,000 0.41 / 0.82
- Dreame L60 Ultra PE 30,000 0.57
- ECOVACS DEEBOT T90 PRO Omni 30,000 1.23
- Roborock Saros 10R 22,000 0.58
- Dyson R1 Nurovi Dry robot vacuum (Light grey/Vinca blue) 21,000 2.78
- Roborock Qrevo Curv 2 Flow 20,000 0.47
- eufy Omni E25 20,000 0.76
- Roborock Qrevo S Pro 18,500 0.31
- Dreame D20 Plus 13,000 0.53
- MOVA P10 Pro Ultra 13,000 1.08
- Roborock Q7 M5+ 10,000 2.24
Claims: MOVA US store, Roborock, Dreame, ECOVACS US, Roborock (Amazon listing title), Dyson, eufy. Measured at the brush: Vacuum Wars.
A bigger suction claim doesn't mean more suction. Of the 12 robots in Swept's home-fit finder with both a maker's claim and a measurement, the one that claims the least, the Roborock Q7 M5+ at 10,000 Pa (opens in a new tab), measured 3.7 times the suction at the brush of the one that claims the most, the MOVA V70 Ultra Complete at 42,000 Pa (opens in a new tab): 2.24 kPa (opens in a new tab) against 0.6 kPa (opens in a new tab), both measured by Vacuum Wars. Across all 12, the rank correlation between the claim and the measurement is -0.13, or +0.03 with the Roborock Saros 20's compare-page figure. A score of 1 would mean a bigger claim always measured bigger; zero means no link at all. We got zero, give or take a rounding error.
So buy on what the robot did, not what the box says. Tell the finder about your floors, pets, long hair, what lives on the floor, whether you want a mop, your tallest threshold, your lowest furniture, your floor area, your privacy line and your price band. It returns the three best fits for your home, each with what it will get wrong, then every robot it ruled out and why. Below it sits the table: 24 exact US models, 17 measures each, every lab's figure kept in its own place.
Home-fit finder
Tell the finder about your home
Answer any question. The results appear here, and the table below follows your answers.
What the data shows
The pascal figure on the box comes from the maker, under conditions the maker chooses and doesn't publish. Vacuum Wars measures with a manometer at the brush housing at the top setting (opens in a new tab), which is the air that actually reaches your rug. I spent my working life building jet engine sections, and a supplier who sent me a performance figure with no test method attached would have had it straight back. Even honest measurement wobbles: Vacuum Wars prints 0.41 kPa (opens in a new tab) for the Saros 20 in its review and 0.82 kPa (opens in a new tab) on its compare page. We show both, and the headline holds either way.
Labs can disagree wildly about one robot's pet hair. On the same US model, the Roborock Saros 10R, and the same kind of floor, Tom's Guide (opens in a new tab) measured it lifting 98% of pet hair from carpet; TechGearLab (opens in a new tab) measured 56%. Neither is wrong. Tom's Guide spread a light load and weighed the bin after one run; TechGearLab worked real fur into the pile. The 10R's split brush gathers hair in its middle instead of passing it all to the bin, so the test decides the score. An average of the two would describe a carpet nobody owns, which is why the table never prints one.
No robot wins everything, and the cards say so. The Roborock Saros 20 climbed a 2.2 in (57 mm) single step at Vacuum Wars, the tallest in the table, yet RTINGS (opens in a new tab) found "you can't rely on it to avoid pet waste." The ECOVACS DEEBOT T90 PRO Omni (under $600) matched the Dreame X60 Max Ultra Complete (under $1,700) on Vacuum Wars' pet-hair test (opens in a new tab), 100% each, and avoided 23 of 24 objects to the X60's (opens in a new tab) 22. And the Dreame L40 Ultra Gen 2 is not the L40 Ultra reviewers once praised: it avoided 2 of 24 objects (opens in a new tab) and let 69% of long hair wrap its brush.
How to read a cell
Each cell leads with one figure, the lab or maker behind it, a marker saying whether it was measured, seen by an owner or claimed by the maker, and the date we checked it, linked to the source. Where the figure isn't confirmed as the exact US model's, the cell says whose it is, most often because the lab's robot shipped in another bundle, paired with a plainer or fancier dock than the one on your listing.
Why do two labs give different numbers for the same robot?
Because they test different things. The exact model's own figure leads, with Vacuum Wars preferred where it measured. Every other lab's figure sits beside it, marked "differs" or "same result", with that lab's method in a line. The page counts the disagreements and never averages them. The stain column is a special case: Vacuum Wars changes its stain scale between reviews, so each figure is shown only against the average its own review stated, and never set beside another robot's.
What if nobody has measured it?
The cell says "not measured". Never a dash, a zero or a guess. Only the exact US model's own results count: a first generation's or a pricier sibling's figure never stands in for it, however alike the names. Where we know why a cell is empty, it says so. The vacuum-only robots have no stain figure because they have no mop.
In what order does the finder list the robots that fit?
First, gates: your price band, your threshold, your furniture gap, mop or no mop, camera, microphone and a Wi-Fi off switch. Each robot passes, passes on the maker's claim only, or is ruled out. A robot with no evidence for a gate is never ruled out; it goes to "Can't confirm from the evidence", with the missing item named.
Then the survivors are scored on Vacuum Wars' figures only, each against Vacuum Wars' own stated average for that test, and only on the measures your answers call for. They're ordered by most measures passed, then fewest claim-only passes, then fewest gaps, then lowest band. No weighted score, no secret sauce. Every card ends with "What it will get wrong": each below-average result and each lab that disagrees, in that lab's words. If accidents happen on your floors, you get every pet-waste result verbatim, because "usually avoids it" is a phrase you'll think about while scrubbing a rug.
One gap sits under every result: no lab measures reliability or support by model. The warranty column is the only comparable figure. That is the thing you'll care about most in year two, and nobody can tell you yet.
Where every number comes from
Every cell is one fact from our research, tied to the exact US model by its model code, with its source, its date and its kind. What stays out on purpose: any price beyond its band, anything drawn from Amazon's customer reviews, Vacuum Wars' composite score (it mixes in review data and price, so it isn't a measurement), and other sites' verdicts on which robot is best. The full method is on how we research robot vacuums.
How to cite it
Can I cite one robot's numbers?
Yes. Every row has a permanent link: this page's address plus # and the robot's id, for example #roborock-q7-m5-plus. The cite box gives the whole table or one row, with the version date from the change log. The CSV download carries every figure with its lab, and every stain figure with its own review's average. The change log, newest first, names every figure that moved.
CC BY 4.0 (opens in a new tab) covers only our own compilation and words: the selection, arrangement, verdicts and notes. Every figure and quote stays its source's; credit the lab or maker named with it.
24 robots · 17 columns · version 2026-11-05
Robot vacuum home-fit evidence table
| Cleaning | Fit | Mopping | Coverage | Suction | Privacy | Support | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Robot | Pet hair picked up from carpet% | Debris dug out of carpet% | Long hair wrapped on the brush% | Objects and pet waste avoidedof 24 | Highest threshold crossedmm | Gap it needs under furniturein | Mop | Dried stains removed (Vacuum Wars score)pointsVacuum Wars scores each review's stain test on its own scale. Read each figure against the average beside it, not against other robots' figures. | Floor covered per chargesq ft | Suction the maker claimsPa | Suction measured at the brushkPa | Camera on board | Microphone on board | Wi-Fi can be switched off | Control without the maker's cloud | Security updates promised, at leastyears | Warrantyyears |
 |
91% | 93% | 0% | 0 of 24 | 20 mm | robot height3.8 in | flat drag pad, no lift, no wash; remove for carpet | 90 points, below this review's average of 107.8 | 2,761 sq ft | 13,000 Pa | 0.53 kPa | No | No | Yes |
No Matter claim; no official Home Assistant integration for Dreame (home-assistant.io/integrations/dreame is 404); not on the Valetudo list. Cloud app only. | Not measured | 3 years |
 |
Not measured | 90% | 69% | 2 of 24 | 20 mm | robot height3.82 in | dual spinning pads, one extends up to 40 mm | Not measured | Not measured | 25,000 Pa | Not measured | No | No | Yes |
Matter promised via OTA (not confirmed). No official Home Assistant integration. Valetudo supports the L40 Ultra (serial starting R2492) but not the Gen 2 by name; the list is exhaustive, so treat the Gen 2 as unsupported. | Not measured | 3 years |
 |
100% | 94% | 0% | 21 of 24 |
19 mm
2 more figures
| robot height4.09 in | dual spinning pads, extending pad and side brush | 85 points, below this review's average of 108 | 1,202 sq ft | 30,000 Pa | 0.57 kPa | Yes | Yes | Yes |
Matter claimed on the US page (no OTA footnote); room selection over Matter not tested by any source found. No official Home Assistant integration; not on the Valetudo list. | Not measured | 3 years |
 |
100% |
89%
1 more figure
| 0% |
22 of 24
1 more figure
|
51 mm
2 more figures
|
lowest gap it got under3.1 in
1 more figure
| dual spinning pads (Omni-Scrub), one extends; pads detach in the dock | 136 points, above this review's average of 93 | 950 sq ft | 35,000 Pa | Not measured | Yes | Yes | Yes |
Matter: maker footnote said "upcoming OTA"; The Hook Up (Apr 2026) ran its seven flagships including this model over Matter/HomeKit with room selection and internet blocked 3 days, but says "almost all", without naming the exception. No official Home Assistant integration; not on the Valetudo list. | Not measured | 3 years |
 |
89% | 90% | 0% | Not measured | Not measured | Not measured | none (vacuum only) | Not measured no mop | 812 sq ft | 21,000 Pa | 2.78 kPa | Not measured | Not measured | Not measured | Not measured | Not measured | Not measured |
 |
100% | 81% | 0% | 23 of 24 |
32 mm
1 more figure
| robot height3.7 in |
roller (OZMO Roller 3.0, 10.6 in / 270 mm, fresh-water rinsed while mopping, 200 RPM) | 157 points, above this review's average of 108 | 1,372 sq ft | 30,000 Pa | 1.23 kPa | Yes | Yes | Yes |
Matter (Vacuum Wars; not on the US product page). Home Assistant: official ECOVACS integration is cloud push (account login), with an option to point it at a self-hosted server instead of the ECOVACS cloud. Not on Valetudo. | Not measured | 1 year |
 |
100% | 89% | 0% | 15 of 24 | 20 mm | robot height4.35 in | roller (HydroJet, self-rinsing) | 119 points, above this review's average of 108 | 1,191 sq ft | 15,000 Pa | Not measured | No | No |
no off switch in the US guide (the two-button 3 s hold only resets Wi-Fi) |
None: no Matter or HomeKit (Alexa and Google only, cloud). No official Home Assistant integration; not on Valetudo. |
none published
Not confirmed as the exact US model'sNo published minimum security-update period found: the US disclosure policy gives fix deadlines only; web searches for an eufy support-period or UK PSTI statement found none. | 1 year |
 |
93%
1 more figure
| 88% | 0% | 21 of 24 | 21 mm | robot height4.4 in | roller (HydroJet, self-rinsing, dual scraper) | 115 points, above this review's average of 110 | 1,070 sq ft | 20,000 Pa | 0.76 kPa | Yes | No |
no off switch in the US guide (the two-button 3 s hold only resets Wi-Fi) |
Matter 1.4 via the eufy Clean app ("Work with Matter"); room cleaning in Apple Home after naming rooms in the app. Offline behaviour of the E25 itself not tested by any source found. |
none published
Not confirmed as the exact US model'sNo published minimum security-update period found: the US disclosure policy gives fix deadlines only; web searches for an eufy support-period or UK PSTI statement found none. | 1 year |
 |
88.5% |
87%
1 more figure
| 0% |
24 of 24
1 more figure
|
32 mm single threshold cleared 1 of 3 attempts
2 more figures
| robot height3.78 in | roller (HydroJet 2.0, extends toward edges) | 82 points, below this review's average of 108 | 1,088 sq ft | 30,000 Pa | Not measured | Yes | No |
no off switch in the US guide (the two-button 3 s hold only resets Wi-Fi) |
Matter (guide section 5); Apple Home room selection and local control with internet blocked (The Hook Up, 3 days). No official Home Assistant integration beyond Matter; one owner says Matter exposes few controls. Not on Valetudo. |
none published
Not confirmed as the exact US model'sNo published minimum security-update period found: the US disclosure policy gives fix deadlines only; web searches for an eufy support-period or UK PSTI statement found none. | 1 year |
 |
29%
Not confirmed as the exact US model'sFlattened (synthetic) pet hair on carpet; average 81%, Eufy C10 100%. Measured on the Roomba 105 Vac + AutoEmpty bundle (Q352020); same 105 Vac robot, different dock. |
82%
Not confirmed as the exact US model'sEmbedded sand in medium-pile carpet; category average 75%, Eufy C10 86%. Measured on the Roomba 105 Vac + AutoEmpty bundle (Q352020); same 105 Vac robot, different dock. |
79%
Not confirmed as the exact US model's7-inch hair tangle test: share of strands wrapped on the brush; average 38%. Measured on the Roomba 105 Vac + AutoEmpty bundle (Q352020); same 105 Vac robot, different dock. |
runs over low objects (cables, small toys); okay with shoes; 'practically nil'1 more figure
|
19 mm
Not confirmed as the exact US model's0.7 in. RTINGS lists both the plain-dock and AutoEmpty versions and does not say which was its test unit. |
lowest gap it got under4.1 in
Not confirmed as the exact US model's10.5 cm. RTINGS lists both the plain-dock and AutoEmpty versions and does not say which was its test unit.2 more figures
| none (vacuum only) | Not measured no mop |
630 sq ft
Not confirmed as the exact US model'sBattery efficiency 0.79 m2 per % battery vs 1.31 average; average coverage 1,015 sq ft. Recharge and resume is built in. Measured on the Roomba 105 Vac + AutoEmpty bundle (Q352020); same 105 Vac robot, different dock. | Not measured |
2.16 kPa
Not confirmed as the exact US model'sAverage 0.83 kPa. Measured on the Roomba 105 Vac + AutoEmpty bundle (Q352020); same 105 Vac robot, different dock. | No | No | no procedure in the owner's guide |
Home Assistant's Roomba integration (local push) "does NOT work with the newer x05 Wi-Fi models, such as Roomba 105". Matter not found for the 105 (iRobot added Matter to the Plus 500 and Max 700 in late 2025). Not on Valetudo. | Not measured | Not measured |
 |
Not measured | Not measured | Not measured |
identified and avoided a cable, a Lego toy and fake pet poop on the hardwood corner test; slowed the side brush near obstacles | 19 mm | lowest gap it got under4.1 in | none (vacuum only) | Not measured no mop | Not measured | 13,000 Pa | Not measured | Not measured | Not measured | Not measured | Not measured | Not measured | Not measured |
 |
Not measured | Not measured | Not measured | read the clutter course and worked around it cleanly | Not measured | Not measured | none (vacuum only) | Not measured no mop | Not measured | 30,000 Pa | Not measured | Not measured | Not measured | Not measured | Not measured | Not measured | Not measured |
| Matic robot vacuum and mop | 100% | Not measured | 7% |
avoided 14 of 24 objects (average 17); felt better in real use, rarely ran over cords2 more figures
| 25 mm | robot height7.8 in | roller (microfiber mop roll) | 147 pointsWell above average while dispensing 0.8 g of water. | Not measured | Not measured | 8.11 kPa | Yes | Yes | not published; Matic says Wi-Fi is required for full function |
Smart-home pairing from the Matic app (Apple Home, Google Home, Home Assistant) over Wi-Fi with IPv6; Home Assistant maps Matic segments to areas for room cleaning; no Thread or Bluetooth pairing. An owner says the link is Matter; Matic is not among the companies in the CSA certified-product filter. | none published | 1 year |
 |
66%
1 more figure
|
81%
2 more figures
| 4% |
19 of 24
3 more figures
|
25 mm
1 more figure
| lowest gap it got under4.1 in |
two mop pads on holders, one extends (MopExtend); lifts 10.5 mm; pads stay on | 73 points, below this review's average of 93 | 2,157 sq ft | 13,000 Pa | 1.08 kPa | Yes | Yes | Yes |
Valetudo supports the "Mova P10 Pro Ultra" (serial starting R2491; "NOT sold as the P10 Ultra"), rooted with the Dreame breakout PCB, warranty seals intact. The US manual file is R2491C, matching that prefix. No Matter claim and no official Home Assistant integration otherwise. |
none published
Not confirmed as the exact US model'sNo security-update period on MOVA's US after-sales policy (warranty, returns, repairs; support@mova-tech.com) or its US privacy policy; the earlier warranty URL now returns 404. MOVA is a Dreame subsidiary (CISA), and Dreame Australia promises 3 years for 'Dreame products', not stated to cover MOVA. | Not measured |
| MOVA V70 Ultra Complete Robot Vacuum and Mop | 94% | 92% | 0% |
23 of 24
1 more figure
|
51 mm
1 more figure
| Not measured |
extending mop (MaxiReachX, 17 cm reach); pads left in the dock for vacuum-only runs | 133 points, above this review's average of 108 | 723 sq ft | 42,000 Pa | 0.6 kPa | Yes | Yes | Yes |
None found: no Matter claim on the US page or in the manual, no official Home Assistant integration, not on the Valetudo list (only MOVA Z500, S20 Ultra, P10 Pro Ultra, V50 Ultra and Z60 Ultra Roller are). |
none published
Not confirmed as the exact US model'sNo security-update period on MOVA's US after-sales policy (warranty, returns, repairs; support@mova-tech.com) or its US privacy policy; the earlier warranty URL now returns 404. MOVA is a Dreame subsidiary (CISA), and Dreame Australia promises 3 years for 'Dreame products', not stated to cover MOVA. | Not measured |
 |
Not measured |
96%
Not confirmed as the exact US model'sJamie Andrews' video does not state the unit's market; same model name (Freo X10 Pro). 54 g coloured rice on medium-pile carpet; ties Freo Pro. | Not measured |
poor: bumped all objects incl. fake pet waste on carpet and hard floor; ignored the cordNot confirmed as the exact US model'sJamie Andrews' video does not state the unit's market; same model name (Freo X10 Pro). | 20 mm | Not measured | two spinning triangular pads, one extends (MopExtend / EdgeSwing) | Not measured | Not measured | 11,000 Pa | Not measured | No | No | no Wi-Fi off procedure in the US manual |
None found: no Matter claim on the US page or in the manual; no official Home Assistant integration (home-assistant.io/integrations/narwal is 404); not on Valetudo. |
2 years
Not confirmed as the exact US model's"at least 2 years from the launch day of products"; the table (updated Apr 2024) lists T10, Freo, Freo X Ultra, Freo Z Ultra and Freo X Plus but not the Freo X10 Pro (YJCC027). All listed end dates before Apr 2026 have passed. | 1 year |
 |
79% | 75% | 43% |
“avoided obstacles more effectively than the others tested, though not perfect”1 more figure
|
20 mm
1 more figure
|
robot height3.9 in
1 more figure
| static drag pad, no lift, no scrubbing; washed by hand | 158 points, above this review's average of 93 | 1,060 sq ft | 10,000 Pa | 2.24 kPa | No | No | Yes |
Home Assistant: Q-series support "varies" (newer Q variants may use a protocol that fails); cloud account needed; no Matter in the app (RTINGS, Q7 M5). Valetudo: not supported (only the older Q7 Max is listed).Not confirmed as the exact US model'sHome Assistant docs (2026.10): Roborock vacuums use local polling with cloud fallback; setup needs the Roborock app account (email code); "Can I block internet access for this device? No" (local API blocks itself without the Roborock servers); over Matter it can run fully offline with limited controls. Valetudo list read 2026-10-26: https://valetudo.cloud/pages/general/supported-robots/ | 2 years | 1 year |
 |
77.5% |
87%
1 more figure
| 0% |
16 of 24
1 more figure
|
19 mm
1 more figure
|
lowest gap it got undernone cleared
1 more figure
|
SpiraFlow roller, 270 mm, rinsed with fresh water in real time; extends for edges | 25 points, below this review's average of 112 | Not measured | 20,000 Pa | 0.47 kPa | Yes |
Yes
1 more figure
| Yes |
Matter: yes, and The Hook Up ran its 2026 flagship group (including this model) over Matter in HomeKit with room selection while blocked from the internet for 3 days ("almost all" worked; the exception is not named). Home Assistant native integration: local with cloud fallback, cannot block internet. Valetudo: not supported. | 2 years | 1 year |
 |
43% | 81% | 62% |
5 of 24
1 more figure
| Not measured | robot height3.8 in | dual spinning pads, 200 RPM max | 149 points, above this review's average of 108 | 1,495 sq ft | 18,500 Pa | 0.31 kPa | No | No | Yes |
Home Assistant: Qrevo series "fully supported" with local polling, but cloud needed for setup and the local API stops without internet. Matter promised by OTA (not confirmed delivered). Valetudo: not supported. | 2 years | 1 year |
 |
92%
3 more figures
| 80% | 0% |
24 of 24
2 more figures
|
38 mmRTINGS's tallest test
2 more figures
|
lowest gap it got under3.1 in
1 more figure
| dual spinning pads, one on an extending arm; auto-detach in the dock | 103 pointsDried stain; 'better than average'; water penalty slightly worse than average (compare page 1.05 g). | Not measured | 22,000 Pa | 0.58 kPa | Yes | Yes | Yes |
Matter local (firmware 02.32.44+). Home Assistant native: Saros "fully supported", local polling but cloud required and local API stops without internet. Valetudo: not supported. | 2 years | 1 year |
 |
88%
2 more figures
|
89%
1 more figure
| 0% |
17 of 24
2 more figures
|
57 mm
4 more figures
|
lowest gap it got under3.1 in
1 more figure
| dual spinning pads, one extends (FlexiArm); auto-detach in the dock | 116 points, above this review's average of 94 | Not measured | 36,000 Pa |
0.41 kPa
1 more figure
| Yes | Yes | Yes |
Matter: yes (maker). The Hook Up ran it over Matter/HomeKit with room selection and internet blocked for 3 days ("almost all" of seven worked; exception unnamed). Home Assistant native: local with cloud fallback; cannot block internet. Valetudo: not supported. | 2 years | 1 year |
 |
Not measured | Not measured | Not measured |
no camera or object recognition; may treat thick carpet or cords as obstacles; tidy before runs1 more figure
| 19 mm | lowest gap it got under4.1 in | none (vacuum only) | Not measured no mop | Not measured | not published | Not measured | No | not listed | Not measured |
None: no Matter (SharkNinja is not among CSA certified-product companies); Home Assistant's Shark IQ integration is cloud polling; not on Valetudo. | none published (US); UK list does not cover the AV25 series | 1 year |
 |
89% | 80% | 54% |
worked effectively only about half the time in the studio; above average for its price, short of flagships | 20 mm | Not measured | none (vacuum only) | Not measured no mop | 731 sq ft | Not measured | 0.6 kPa | Not measured | Not measured | Not measured | Not measured | Not measured | 1 year |
 |
40%
1 more figure
| Not measured | Not measured |
runs over flat objects such as socks and cords; pushes small items several feet; trips on thin cords and chair legs | 22 mm | robot height3.9 in |
clip-on static pad under a 300 ml tank; electric water-flow control, 3 levels; no pad lift | Not measured | Not measured | 5,300 Pa | Not measured | No | No |
no Wi-Fi off for cleaning; Energy Saving mode (hold Dock 15 s) disconnects Wi-Fi but leaves only charging |
Matter 1.3 (firmware 1.2.1); Matter 1.3 predates the 1.4 Service Area cluster used for room selection, so room selection over Matter is not expected (not tested). Home Assistant TP-Link integration controls Tapo vacuums locally "without going via the TP-Link cloud" (lists RV30 Max and RV20 Max Plus), but needs the TP-Link cloud username and password to authenticate. Not on Valetudo.Not confirmed as the exact US model'sHA lists "RV30 Max", the same robot sold without the self-empty dock; not tested on the Max Plus bundle. Matter firmware: https://www.tp-link.com/us/support/download/tapo-rv30-max-plus/ | Not measured | 1 year |
Licence: CC BY 4.0 (opens in a new tab) covers only our own compilation and words: the selection, arrangement, verdicts and notes. Every figure and quote stays its source's; credit the lab or maker named with it.
Cite this
Swept, Robot vacuum home-fit evidence table, version 2026-11-05, https://sweptfloors.com/robot-vacuum-home-fit-finder/. CC BY 4.0 covers only our own compilation and words: the selection, arrangement, verdicts and notes. Every figure and quote stays its source's; credit the lab or maker named with it.
Change logversion 2026-11-05
Newest first. Each entry names every row and column whose lead value, lab or kind changed, every cell whose other entries changed, and every row added or retired. Counts are taken from the data, never typed.
2026-11-05 (re-audit)
- Notes keep to what each lab measured. Three notes no longer say how another lab or site placed a robot, or give another robot's result (the X60 and Saros 20 threshold notes, the Shark mop note). No figure, lead, lab or kind changed (319 filled cells, unchanged).
- One more note keeps to the measurement. The L60 Ultra PE's carpet note gives its result and the lab's average only. No figure changed.
- Tom's Guide's eufy C28 and S2 tests are the UK-team unit's. The C28's overall and cat-litter results and the S2's overall result were filed as the US models'; they are now marked as the UK unit's, like that review's pet-hair figures already were, and never lead a cell.
2026-11-04: suction figures added, exact-variant figures lead, and other models' figures leave
Sixth rebuild: other markets' support dates leave
- Six cells emptied, security updates: Dreame D20 Plus, L40 Ultra Gen 2, L60 Ultra PE and X60 Max Ultra Complete (all led by Dreame's Australian site's 3-year statement), ECOVACS Deebot T90 Pro Omni (ECOVACS's UK/Australia policy) and iRobot Roomba 105 Vac (iRobot's UK statement). Each registry fact now names the market of what it covers, and another market's statement is never the US robot's, so each cell shows "not measured" (filled cells 319 of 408, was 325). No other cell changed.
Fifth rebuild: the Tapo's UK support date leaves
- One cell emptied: Tapo RV30 Max Plus, security updates. Its only entry was TP-Link's UK PSTI statement ("2029-12-31 or later"), which covers UK hardware V1; the US robot is hardware V1.8 and has no published date. Another market's statement is never the US robot's, so the cell now shows "not measured" (filled cells 325 of 408, was 326). No other cell changed.
Fourth rebuild: each stain figure read against its own review's average
- Dried stains removed: each Vacuum Wars figure now carries the average stated in its own review and is judged above, at or below that average only, never against one fixed bar or a figure from another review (13 of the 15 cells carry a stated average; the Matic's and the Saros 10R's reviews give theirs only in words, which the cells show, unjudged). No lead value, lab or kind changed.
- eufy Omni S2, dried stains removed (Vacuum Wars): the note no longer quotes the S1's figure, another model's result.
- Roborock Saros 10R, objects avoided (TechGearLab): the note again says the robot was one of two to clear that test's mark; it states the test's outcome, not a ranking (figures beside the leads: 54, was 53).
Third rebuild: the remaining other-market figures and reviewers' superlatives leave
- One cell emptied: MOVA V70 Ultra Complete, gap under furniture. Its only figure, TechRadar's 3.4 in, was measured on a UK unit; the cell now shows "not measured" (filled cells 326 of 408, was 327).
- Two leads lost a reviewer's superlative and kept what was seen: iRobot Roomba Max 705 Vac and Max 715 Vac, objects avoided (Just A Dad Approved): the observation stays, the comparison with the reviewer's other Roombas and the score out of ten go.
- Another market's figures left the cells' other entries: ECOVACS Deebot T90 Pro Omni, threshold (Cybershack) and objects avoided (Notebookcheck), Australian and European units; MOVA V70 Ultra Complete, pet hair (TechRadar, Tom's Guide), UK units; Dreame X60 Max Ultra Complete, pet hair (Tom's Guide), a UK unit; Narwal Freo X10 Pro, objects avoided (Basic Tutorials).
- Reviewers' superlatives and awards left the notes (named by row, column and lab, never quoted): Matic, suction measured and robot height (Vacuum Wars), objects avoided (RTINGS); Roborock Saros 10R, objects avoided (Vacuum Wars, TechGearLab); Dreame L40 Ultra Gen 2, objects avoided (Vacuum Wars), which also lost a sentence about the first-generation model; eufy Omni S2, carpet (The Hook Up).
- Method notes keep their separators ("floor: carpet | mode: two runs"), which the second rebuild had run together.
- Figures beside the leads: 53 (was 60).
Second rebuild: other models' figures, other sites' rankings and human-hair results leave the table
- No cell changed its lead, and no cell was emptied: every robot shows the same leading figure as in the first rebuild below. What left are figures that sat beside the leads, and sentences in the notes.
- Another model's figure left the table and the CSV entirely, instead of sitting beside the exact model's own figure:
- Dreame L40 Ultra Gen 2, gap under furniture: RTINGS's 4.1 in, measured on the first-generation L40 Ultra. Dreame's 3.82 in for the Gen 2 still leads.
- eufy Omni C28, pet hair: Tom's Guide's 82% (hardwood) and 92% (low-pile carpet), and Expert Reviews' 98%, all on UK units.
- eufy Omni S2, pet hair: Tom's Guide's 86% (hardwood) and 82% (low-pile carpet), on a UK unit.
- MOVA V70 Ultra Complete, pet hair: Tom's Guide's 84% (hardwood), on a UK unit.
- eufy Omni E25, debris dug out of carpet: SmarthomeAssistent's 61.43%, on a European Omni E28.
- eufy Omni S2, debris dug out of carpet: SmarthomeAssistent's 69.73%, on a European unit.
- MOVA V70 Ultra Complete, objects avoided: PC-WELT's result, on a European unit.
- Human-hair results left the pet-hair column. SmarthomeAssistent, a German lab, tests with human hair on European units, and its 22 results had been filed as pet-hair pickup. They are now recorded as human-hair pickup, which is not a column here. They left the pet-hair cells of seven robots:
- Dreame X60 Max Ultra Complete: 91%, 87% and 99%, plus two results of 100% that had agreed with the lead;
- ECOVACS Deebot T90 Pro Omni: 99%, plus two results of 100% that had agreed with the lead;
- eufy Omni E25 (tested as the E28): 100% twice;
- eufy Omni S2: 100% twice, and 61%;
- MOVA V70 Ultra Complete: 100%, 78% and 82%;
- Roborock Saros 10R: 100%, 85% and 75%;
- Roborock Saros 20: 100% twice, and 99%.
- Other sites' picks and rankings left the notes. The figures stay; only the sentence ranking the robot went, from these notes:
- Dreame L60 Ultra PE, debris dug out of carpet (Vacuum Wars): its place among the robots Vacuum Wars has tested;
- eufy Omni S2, objects avoided (Vacuum Wars): its place among the robots Vacuum Wars has run through its obstacle course;
- eufy Omni E25, pet hair (TechGearLab): its place among the robots TechGearLab tested;
- Roborock Q7 M5+, objects avoided (Wirecutter): Wirecutter's pick for it. The cell's lead is still Wirecutter's own description of how it avoided obstacles, shown as Wirecutter's words, with RTINGS's result beside it;
- Roborock Q7 M5+, suction measured (Vacuum Wars): its place among the robots Vacuum Wars has measured.
- Counts: 327 of 408 cells filled (unchanged); 60 entries differ from their cell's lead (was 88) and 4 agree (was 8), in 17 of 24 rows (was 19). By column: objects avoided 19 (was 20), highest threshold 18 (unchanged), pet hair 11 (was 35), carpet debris 6 (was 8), furniture gap 5 (was 6), suction measured 1 (unchanged). Pet hair still covers 19 of 24 robots (79%).
- The headline (suction claim against measurement) and the pet-hair second fact were rechecked by their rules on this data, and both still hold.
First rebuild: suction figures added, and exact-variant figures lead
- MOVA V70 Ultra Complete, suction measured: now 0.6 kPa (Vacuum Wars, compare page). Was not measured.
- Roborock Qrevo S Pro, suction measured: now 0.31 kPa (Vacuum Wars, compare page). Was not measured.
- Roborock Saros 20, suction measured: Vacuum Wars' compare-page figure, 0.82 kPa, now sits beside the 0.41 kPa from its written review, which still leads. Both are Vacuum Wars' own figures for the same robot, shown as a conflict.
- These three figures were in sources kept on 26 October but missing from the table until today.
- Dreame L40 Ultra Gen 2, gap under furniture: now led by Dreame's own figure for the Gen 2, a 3.82 in robot height (maker's claim). RTINGS's 4.1 in, measured on the first-generation L40 Ultra, sits beside it, labelled as that model's.
- iRobot Roomba 105 Vac, objects avoided: now led by Just A Dad Approved's test of this exact model (it runs over low objects such as cables and small toys). Vacuum Wars' result, measured on the AutoEmpty bundle, sits beside it.
- Roborock Q7 M5+, highest threshold crossed: now led by Roborock's 20 mm claim for the M5+. RTINGS's 19 mm, measured on the dockless Q7 M5, sits beside it.
- Roborock Q7 M5+, gap under furniture: now led by Roborock's 3.9 in robot height (maker's claim). RTINGS's 3.9 in, measured on the dockless Q7 M5, agrees.
- The last four changed because each cell now leads with the exact model's own figure before a sibling's.
- Notes: the eufy Omni E25 and Narwal Freo X10 Pro warranty notes no longer quote the price of an extended warranty.
- Counts: 327 of 408 cells filled (was 325); 88 entries differ from their cell's lead (was 87) and 8 agree; suction measured covers 15 of 24 robots (was 13).
- Corrected: the first version's entry said 326 filled cells; that version had 325.
2026-10-30: first version
- 24 robots (23 current, plus the Roomba 105 Vac, status unclear) × 17 columns, built from the registry of 27 exact US variants researched 26–28 October 2026.
- 325 filled cells; 87 results that differ from their cell's lead and 8 that agree, shown beside it, never averaged.
- Added to the registry for this table: mop hardware for 10 robots (6 of them vacuum-only), RTINGS's inch furniture-gap figures for 8, robot heights for 3, warranty years for 2 and one threshold claim. Each is from a source already kept, quoted in its note.
Where to go next
The finder narrows the field. Our buying pages make the call:
- A dog or cat that sheds: best robot vacuum for pet hair
- Pet accidents, cords and toys: the best robot vacuum that avoids dog poop, cords and toys
- Hard floors you want mopped: best robot vacuum and mop for hard floors
- No mop wanted: best robot vacuum without mop
- Carpet, high pile and rugs: best robot vacuum for carpet
- Long hair in the house: best robot vacuum for long hair
- No camera or microphone: best robot vacuum without camera or microphone
- A firm budget: under $300, under $600, under $1,000
Single-robot reviews, each linked from its row: Dreame X60 Max Ultra Complete, Roborock Saros 20, Roborock Saros 10R, ECOVACS DEEBOT T90 Pro Omni, eufy Omni S2, Dreame D20 Plus, Roborock Q7 M5+ and Tapo RV30 Max Plus.
The box tells you what a robot hopes to be. The finder tells you, before it arrives, what it's going to get wrong in your house.
The evidence in full
13 sources
Sources
- Vacuum Wars: Roborock Q7 M5 robot vacuum review (opens in a new tab)
- Vacuum Wars: MOVA V70 Ultra Complete compare page (opens in a new tab)
- Vacuum Wars: Roborock Saros 20 review (opens in a new tab)
- Vacuum Wars: Roborock Saros 20 compare page (opens in a new tab)
- RTINGS: Roborock Saros 20 review (opens in a new tab)
- Tom's Guide: Roborock Saros 10R review (opens in a new tab)
- TechGearLab: Roborock Saros 10R review (opens in a new tab)
- Vacuum Wars: ECOVACS DEEBOT T90 Pro Omni review (opens in a new tab)
- Vacuum Wars: Dreame X60 Max Ultra Complete review (opens in a new tab)
- Vacuum Wars: Dreame L40 Ultra Gen 2 video review (opens in a new tab)
- Roborock: Q7 M5+ product page (opens in a new tab)
- MOVA: V70 Ultra Complete product page (opens in a new tab)
- Vacuum Wars: How Vacuum Wars tests robot vacuums (opens in a new tab)