We earn a commission if you buy through the links on this page, at no extra cost to you. It never changes a ranking, and there are products recommended here with no affiliate link behind them at all. Full disclosure.
First: is this tank for a betta?
If it is, we would rather you spent the filter money on a bigger tank. Three rated gallons is 2.6 gallons of actual water, and that is a buffer so small that temperature and ammonia both swing faster than you can react to either. Our floor is five gallons, and it is not a close call.
A 3-gallon is a genuinely good shrimp or snail tank, though, and those need filtration too. So does a quarantine or grow-out tank. The arithmetic below stands either way, and so do the picks - we have simply told you which fish we would not put in front of them.
The sum, for this exact tank
2.6 gal of actual water x 4 turnovers an hour is about 10.4 GPH. Past roughly 15.6 GPH you are baffling the outflow or buying a different filter.
A rated 3-gallon holds about 2.6 gallons once there is substrate in it and you have stopped filling below the rim. The general aquarium convention is four to six turnovers an hour; for a betta we size to 4 and treat anything the flow control cannot bring under 6 as a baffling job rather than a purchase.
| Tank (rated) | Actual water | Target flow (4x) | Ceiling (6x) |
|---|
| 3 US gal | 2.6 gal | 10.4 GPH | 15.6 GPH |
So the target is about 10.4 GPH with a ceiling around 15.6. Ten gallons per hour is a very small number, and it is smaller than almost anything in the filter aisle is built to deliver.
The finding, before the picks
This is the one page on this site where a manufacturer’s published flow rate exists - and it makes the problem worse rather than better.
Tetra publishes 27 GPH for its small internal filter, rates it for 1 to 4 gallon aquariums, and then states outright that it “cycles a 4 gallon aquarium more than 6 times per hour.” That is a manufacturer volunteering that its own product runs at the top of the conventional range. In 2.6 gallons of actual water, 27 GPH is roughly ten turnovers an hour - about two and a half times our ceiling.
Everything else in this class publishes nothing at all. Fluval rates its Nano filter by tank size only. Sponge filters have no manufacturer, let alone a specification. We record each absence on the product entries below, and the single most useful conclusion on this page is that in a 3-gallon you are not shopping for flow - you are shopping for something whose flow you can reduce.
What we ranked these on
We have not run these filters. No test bench, no flow meter, no drawer of samples - so every criterion below is either read off a published page or is a structural property anyone can verify by looking at the product.
Can the flow be brought down? The criterion that decides this page. At 15.6 GPH of headroom, a filter you cannot slow is a filter you will be baffling on day one.
Does the published range actually include 3 gallons? Several popular nano filters are rated from 5 or 10 gallons upward. Buying below a stated range is the same error as buying above it, and it is the reason the most recommended nano hang-on-back is absent here.
Published spec depth. Flow rate, rated range and media contents all present would score 9 or 10. One product here publishes two of the three, which is unusually good for this category.
Footprint inside the tank. In 2.6 gallons, an in-tank filter is a meaningful fraction of the display. This matters far more here than at ten or twenty gallons.
Running cost. Cartridge filters bill you monthly and reset your bacteria each time. Foam rinses in old tank water and goes back in.
Why the AquaClear 20 is not on this page
Because Fluval’s published range for it starts at 5 gallons, and there is no honest way to read that as covering a 3. It is the filter we recommend most often on this site - our review says so - and it is still the wrong purchase here.
The same applies to every hang-on-back rated “5 to 20” or “10 to 30”. Being below a stated range is not a safe direction to err in: the minimum figure exists because the pump has a minimum output, and in a tank this size that output is the entire problem.
The three shapes that do fit
Sponge filter. Air rises through a lift tube and drags water through foam. There is no impeller, so it cannot shred a fin or injure a shrimplet, and the flow is set by an air valve - which makes it the only filtration in this class with genuinely continuous, fine adjustment. In a 3-gallon that adjustability is worth more than any published figure, because it is how you get to 10.4 GPH from a product that never told you what it moves.
Small in-tank power filter. The Tetra clips to the glass and is the only pick here whose published range genuinely includes a 3-gallon. What you are accepting is a fixed 27 GPH with no stated flow control, which means the baffle is not optional for a long-finned fish.
In-tank foam filter with a spray bar. The Fluval Nano is rated to 15 gallons - well above this tank - but it ships with an adjustable output and a spray bar, which is the part that turns a jet into a sheet. It is the gentlest of the three out of the box and the bulkiest in the display.
At three gallons, the baffle is part of the purchase
Budget ten minutes and a scrap of coarse filter foam. Wedged over an outflow it spreads a jet into a sheet and drops the current a fish actually feels, without restricting the intake - never restrict the intake, because that starves the media of the oxygenated flow the bacteria need. How to baffle a filter properly covers both the foam method and the bottle method. On this page it is not a fix for a bad choice; it is a normal part of running any powered filter in a tank this small.
The cartridge problem, which bites hardest here
Small in-tank filters are almost universally cartridge-driven, and a cartridge is a slab of floss wrapped around carbon on a plastic frame. Replacing it monthly - as the box instructs - throws away your entire mechanical layer and most of your bacterial colony at the same time.
In 2.6 gallons that reset is far more dangerous than it would be in a twenty. There is no spare volume to dilute the ammonia that appears while the colony rebuilds. The fix is the same as everywhere else on this site: cut the floss off the frame and pack the space with foam you rinse instead of replace. What each media layer is actually for works through it.
The sponge filter sidesteps the problem entirely, which is a large part of why it ranks first. There is no cartridge to buy, ever.
Setting the flow, since only one of these tells you anything
Fill the tank, run the filter with no livestock in it, and watch the surface. You want a moving surface without a standing ripple pattern, and plants that sway rather than stream. In a tank this small the difference between those two states is a quarter-turn of an air valve.
Then add the occupants and watch where they choose to be. A betta wedged behind hardscape at the far end all day is telling you the flow is too high; shrimp that stay under the sponge rather than grazing the glass are telling you the same thing. That is behavioral evidence rather than measurement, and we would rather label it as such than dress it up.
The filter is not what cycles the tank
A new filter is a box of clean plastic. The bacteria that process ammonia live on the media and take weeks to establish, and in 2.6 gallons there is no margin for getting that wrong. Cycle the tank first and confirm it with a test kit rather than the calendar - a 2 ppm ammonia dose processing to zero ammonia and zero nitrite in 24 hours.
If you are moving down from a larger tank, carry the old media across rather than starting again. It is the single biggest favor you can do a small tank, and it costs nothing.
What we did not rank, and why
Hang-on-backs rated from 5 gallons up. Covered above. Their own published ranges exclude this tank.
Undergravel filters. They work, and in a tank you will be vacuuming weekly with only an inch of substrate they make maintenance considerably worse for no benefit a sponge does not already provide.
Internal filters from rotating brand names. Same standard we apply to heaters: a filter we cannot trace to a manufacturer specification is one we cannot recommend on the basis this site uses. That excludes a large part of what is actually on the shelf at this size, and we would rather have a short list we can defend.
What we cannot tell you
How loud any of these are on a desk - and at three gallons the tank is almost certainly on a desk. Whether the Tetra’s published 27 GPH holds as the cartridge loads up. How the sponge brands differ, given that category has no manufacturer and no published specifications of any kind. Those absences are recorded on the product entries below rather than filled with a guess, which is the standard this whole site works to.
