The 3i S10 Ultra Completely Blew My Mind

The S10 tower is a water factory, for good and ill

3i S10 Ultra tower weighs 50lbs and is huge, but does more than most towers Credit: Amanda Blum

Let’s get this out of the way. There is nothing dainty or streamlined about the S10. First, it arrives in the biggest box I’ve ever seen for a robovac, which is impressive, since robovac boxes are already usually pretty imposing. But it’s the weight that will send you reeling. At 50ish pounds for only the tower, you’ll want to be deliberate about where you place it, because it lives there now and you will never want to move it. The tower isn’t simply heavier than most; it’s also bulky and lacking in the sophisticated edges and clean facades of modern, premium robot towers. 

All of that said, it’s worth it, because the tower has a neat trick: it generates its own water. It does this in two ways—by pulling from the air, and by using an almost boiling hot distillation process to purify and recycle dirty water. In combination, these make the mop on your robot entirely autonomous. No more hauling containers to dump dirty water or top up the clean water. You don’t have to tie the plumbing into your home’s water system, too, since it’s all handled in the tower. According to 3i, the tower can produce 300ml of water from the air per hour, but this is dependent on humidity levels. Still, this is more than enough to keep the water tank full for almost any family.

Not just a gimmick: The cleaning process actually works

Now, if the recycling skeeves you out (skeeves being the technical term), don’t worry. Almost all the water we use has gone through a cleaning process, and this one heats the water to the point of evaporation to distill it before it's recycled. Any solid waste left behind is then moved to the dust bag. You will have to occasionally change the dust bag, but like other robots, it’s occasional. 

The 3i S10 ultra mopping up a cup of dirt Credit: Amanda Blum

That all sounds great in theory, but how legitimate are this wonder robot's claims in practice? After all, a water-making machine sounds a bit like magic. Plus, how clean could it really get the water if it only ever hits sub-boiling temperatures? To test how clean water really got, I pulled a sample from the water tank and tested the PH. It was clear to the eye, and the PH registered as normal, the same as my tap water. Then, I threw pickling liquid, which has a very low PH, on the floor, and let the mop clean it. In the same run, I threw one cup of dirt and water on the floor, and let the mop clean that as well. Six hours later, I tested the water in the tank again. The water had the same PH as it did that morning and there was no dirt present. It looked crystal clear to the eye.

On the left, water taken from the tank after mopping up low PH liquid and a cup of dirt. On the right, the water from the tank before doing the cleaning. They have the same PH, and appear to be crystal clear. Credit: Amanda Blum

But what about your power bills? All this water distilling and dehumidifying would be concerning, from an electrical usage standpoint. I haven't had the robot long enough to track these stats over time, but according to 3i, the robot uses just over 1.5 kWh per day. While that is higher than many other robots I've tested (the Roborock 10R comes in under .5 kWh), it's not outlandish. The Dreame X50 Ultra clocks in just under the 3i.

Instructions could be better, both in print and in the app

Instruction diagrams are not terrifically helpful Credit: Amanda Blum

I do have a few complaints about the 3i, and I encountered one during the unboxing and assembly phase. The instructions kind of suck. The robot comes with an instruction book that was so slight I believed it did not exist and had to revisit the box to find it. But even so, the diagrams on it are not particularly helpful. There’s also a sticker on the top of the tower to help you, too, but the diagrams there are the same. As such, I installed the cleaning fluid upside down and backwards (a real feat) the first time I set it up. Still, it’s a robovac, and even without any instructions whatsoever, you’d quickly figure out how to use this one. Scan a QR code, install an app, follow the instructions the app offers. That part took less than five minutes. 

The 3i app is fanciful, and you can watch animations of the robot being cleaned there, but I thought the organization of the functions could be better. I often had to go back and forth between screens to find the settings I was looking for. For instance, as soon as you enter the app, you have some “quick commands” that feel less like shortcuts and more like settings: deep clean, full sweep, AI auto, and full vac and mop. These will send your robot out, using these presets on all your rooms. To set up a cleaning with more specific preferences for which rooms to clean, as well as what vacuum and mop strength to use, you need to click “access device”.  

Settings for a run are buried three screens deep, and not intuitive. Credit: Amanda Blum

Once in the right screen to create those custom runs, I found the app to be not highly responsive or intuitive. It takes a number of taps to get the room or zone to respond, and then you have to go to another screen to tell the robot what settings to use. Once, I found the settings to tell the tower to manually empty the robot, but I was never able to uncover them again. The app is clunky, and could get smoother, but ultimately still works and isn’t so annoying as to be a deal breaker. Software can be updated. 

This robot ate and left no crumbs

The front sweeps brush everything into the path of the robot. Credit: Amanda Blum

I’ve spent a lot of time pining over the updated sweepers on various premium vacuums this year, like the Roborock, Dreame, and Eureka. Companies have spent a lot of time and thought deciding how many sweepers there should be on a robovac, how many brushes they should have, and how far out the arms they’re on should extend. I actually laughed at how elementary the sweepers on the S10 Ultra are by comparison. Two gigantic, three armed sweepers positioned on the front corners of the robot rotate in opposite directions. There are no extending arms here, and the sweepers are instead so large they extend out beyond the robot, with their brushes pushing against the floor.

This simple approach works, though. The sweepers effectively direct all debris right into the mouth of the robot. And when i say effectively, I mean this robot eats. I have never, ever seen a robot that can suck up the debris the S10 can, and do so without needing the roller cleared. The S10 didn’t even have to return to the base that often. I watched it suck up gigantic mulch sticks. Dog toy floof. Leaves, labels.  I started throwing random crap in its path because it was legitimately awe inspiring.

It’s not just large debris, either. The S10 uses a green laser light to highlight the debris in front of it, and the sweepers do an exceptional job getting small dog hair, microdust, pretty much everything into the path of the robot. The sweepers also dig very effectively under baseboards, floor furniture and the edges of rugs to clean, too. More so than any other robot I’ve experienced. My area rug was cleaner than any other robot had achieved before. The carpet had lovely tracks and no visible debris was left behind.

Impressive results from nothing special stats

But here is what makes my great experience even more mind blowing. The S10 Ultra is doing all of this using only 13,000Pa of suction power, and doing it quieter than any other vacuum I’ve tested. Let me dig in there. Recently, I made mention of the fact that the suction power on robovacs is going through the roof, having doubled since last year on most models. The Dreame x50 is 20,000Pa, the Roborock Saros 10 is 22,000Pa, and the Ecovacs x8 is 18,000Pa. 13,000Pa? That’s nothing, comparatively, and yet my beloved Roborocks get caught up on the large debris the 3i handled with ease. And no matter what the robot sucked up, it never made the rattling sound so many other robots do, and never got louder. Even on max suction and mopping settings, it was easy to hear the television over. 

The S10 Ultra got into small spaces Credit: Amanda Blum

The robot was also moderately adventurous in terms of getting into smaller spaces. A few weeks ago, I noted that the Roborock Saros 10R, which relies solely on AI and cameras and had ditched LiDAR (a navigation system based on light that is employed by almost all modern robovacs), was willing to go into nooks and crannies. I chalked it up to ditching the LiDAR. The S10 Ultra is still very much LiDAR dependent, even though it adds AI and has on board cameras, as well (most premium robots have both these days as they transition to camera only), but was still willing to navigate the exterior of the space it was cleaning and get up close to what it could. I was also impressed at how well the robot fit under my furniture, because it certainly appears to be taller than competitors (which are trying to attain slimmer and lighter robots). I have to point out that while it always ultimately surmounted the thresholds in my home, it was clearly a struggle for it to do so at times. The S10 Ultra is almost 4.5 inches tall and also weighs a lot more than its competitors. If this is what it takes to get fantastic cleaning from a robot, though, I’ll take it. 

Better than other mops, but still room for improvement

Again, much thought is spent at most robot headquarters deciding between a mopping pad or spinning mop pads, but here, too, the 3i has deviated and gone with a mop roller. The roller is cleaned by a line of jets as the robot moves across the floor, and then a scraper squeegees the dirty water off, so theoretically, dirty water is never pushed around the floor. Cleaning fluid is automatically added to the water via the tower. As a side note, at almost $40 a bottle (yes, you have to buy the proprietary bottle) I was surprised that the only two consumables that aren’t tracked for replacement notifications here are the cleaner and the vacuum bag (arguably, two of the most important components). 

The S10 mop was effective and performed as well or better than any other premium robot I’ve tested. Still, it wasn’t perfect (no robot mop is), and had trouble with grime and floor surfaces that weren’t smooth. For instance, on smooth tile, it clearly struggled with corners that were slightly indented, and with grout lines. On tile that wasn’t smooth, it also wasn’t able to get the tiles sparkling, and the pitted areas still had ground in dirt. The robot seemed to handle wood floors without problem, though.

S10 Ultra roller extends to the wall Credit: Amanda Blum

Although the mop is a roller, it was also still able to extend out to meet the wall or get closer to objects than the robot could. However, for reasons I can’t explain, this feature is only available when the S10 is in vacuum and mop mode (as opposed to vacuum, then mop, or only vacuum or only mop).