Press-Fit Tolerances for 3D Printing: Numbers That Survive Contact With Reality

Somewhere in your CAD history is a part with a hole “exactly” the size of the pin that goes in it. You know how that print ended: either the pin rattled around like a pencil in a mug, or you leaned on it with both thumbs, heard a crack, and printed the part again.

Fits are the most re-printed feature in 3D printing, and the standard advice — add 0.2mm of clearance — is half an answer. It works for some people because their printer, material, and intended fit happen to match the person who wrote it. Then you change one of those three things and the folk number stops working.

This guide separates the two questions that “add 0.2mm” smashes together, gives you the FDM fit table with sources, and shows the 10-minute test that makes any of it exact for your machine. The Press-Fit Calculator runs all of this live.

Two questions, not one

Question 1: What gap should the finished parts have? This is mechanical design, independent of 3D printing. A bearing pressed into a pocket wants slight interference. A hinge pin wants a bit of air. Machinists solved this a century ago (ISO 286 fit tables) — but their numbers don’t transfer directly, because…

Question 2: What do I model so the printed parts actually measure that? FDM parts miss their modeled dimensions in predictable directions: holes print undersize, external diameters print slightly oversize. If you don’t correct for both, your designed clearance is fiction.

Most advice bakes both into one number. Keep them separate and fits become boring and repeatable.

The FDM fit ladder

Designed gap between finished parts, on diameter, for rigid filaments (PLA, PETG, ABS) in the 5–25mm range:

Fit	Designed gap (diametral)	Feels like	Use for
Press / interference	−0.10mm	Firm thumb or vise pressure, stays put	Bearings, pins, bushings, magnets in pockets
Snug / transition	+0.05mm	Slides with light friction, holds position	Locating features, lids that shouldn’t fall off
Close running	+0.15mm	Moves freely, no visible play	Pivots, printed hinges, sliding mechanisms
Free / loose	+0.35mm	Obvious air, never binds	Outdoor parts, dusty environments, “must work first try”

Cross-checking these against published numbers: Markforged’s printed unit tests found ~0.05mm of interference “just right” for press fits on their industrial machines (tighter machines, tighter fits); community engineering guides land at −0.1 to −0.2mm interference for desktop FDM; Prusa’s design guidelines call for “at least 0.3mm” of modeled clearance on anything that must move; and Hubs’ snap-fit guide recommends a full 0.5mm of modeled clearance between FDM mating parts. Our ladder plus the print compensation below reproduces those modeled values almost exactly — which is the point: the folk numbers you’ve seen are fit-plus-compensation, pre-mixed for someone else’s printer.

Why is FDM interference so much bigger than machinist interference? An ISO H7/p6 metal press fit at 20mm is ~0.02-0.04mm of interference — plastic uses 0.1mm+. Metal barely stretches; plastic’s low elastic modulus means it deforms generously and grips anyway. That elasticity is also your error budget: it’s why press fits work at all on a machine that holds ±0.1mm on a good day.

Why holes and shafts miss in opposite directions

A printed hole comes out 0.1–0.3mm undersize for three compounding reasons — slicer polygon approximation, extruder overshoot on inside curves, and cooling shrinkage. The full physics with per-material numbers is in our holes guide; the punchline is ~0.24mm undersize for a 5mm hole in PLA on a 0.4mm nozzle.

A printed shaft comes out ~0.1mm oversize (community measurements run 0.1–0.2mm). Same extrusion physics, opposite sign: material that piles up on the inside of a curve narrows a hole but widens a pin. There’s no polygon undersizing on external perimeters, so shafts err less than holes — roughly half as much.

Stack them and the trap is visible: model a 6.0mm pin and a 6.1mm hole (“0.1mm clearance!”) and you actually print a ~6.1mm pin in a ~5.9mm hole. That’s an interference fit you designed as a running fit. Everyone who’s fought a print-in-place hinge that fused solid has met this trap.

Both compensations plus your chosen fit gap, applied to the right sides:

hole in CAD  = nominal + fit gap + hole compensation (+0.24mm-ish)
shaft in CAD = nominal − shaft swell (−0.10mm-ish)

The numbers in practice

608 bearing (22mm OD) pressed into PLA: model the pocket at ~21.95mm plus your printer’s hole compensation — practical guides converge on 0.05–0.10mm of interference per side for PLA, up to 0.10 for PETG. Add a 0.5mm × 45° lead-in chamfer and press with a vise, not a hammer; the chamfer does more for success rate than another 0.05mm of anything.

8mm printed hinge pin in a printed hole: pin at 7.90 in CAD, hole at 8.15 + 0.24 ≈ 8.39. Prints as a real ~8.0 pin in a real ~8.15 hole. Free movement, no slop.

Two printed boxes that slide together: +0.15mm per mating pair if you want quality drawer-feel; +0.3mm if it lives outside or you’re printing PETG (whose slightly tacky surface eats ~0.05mm of any running clearance — engineering tolerance guides call this out specifically).

TPU anything: different world. TPU compresses and grips — press fits want double the interference, and moving fits need 0.5–0.8mm of total clearance or they weld themselves shut with friction.

Three tricks the tables don’t tell you

1. Put the precision on the shaft. External dimensions print more accurately, and you can measure a shaft in two seconds with calipers — then sand it if needed. You can’t sand the inside of an 8mm hole. When you control both parts, design the hole loose-ish and tune the pin.

2. Crush ribs beat round interference. For press fits that need to survive many insertions, AON3D’s assembly guide recommends modeling three or four small ribs (~0.2mm proud) inside the hole instead of shrinking the whole diameter. The ribs crush on insertion, the fit self-centers, and the hole’s walls don’t split — round interference holes concentrate hoop stress and crack along layer lines.

3. Plastics creep. A PLA press fit under constant stress relaxes over months, faster anywhere warm. For load-bearing permanent joints, add a mechanical backup — a shoulder, a groove and snap ring, a drop of CA — or use a bolt with a heat-set insert instead of trusting friction forever.

The 10-minute coupon

No table beats one test print, and you don’t test with the real part:

1. Print a 10×10mm plate containing your hole, and a 10mm stub of your shaft. Minutes, not hours.
2. Test the fit. Wrong? Adjust the offending dimension 0.05mm and reprint the coupon.
3. Two iterations maximum. Write the working numbers down — they stay valid for that printer + material combo for months.

A calibrated peer-reviewed study of FDM fits (Processes, 2023) found desktop PLA machines hold IT11–IT13 tolerance grades after calibration — good enough for real clearance fits, which is exactly what the coupon gets you without the lab.

Run your numbers in the Press-Fit Calculator — it applies the ladder, both compensations, material adjustments, and orientation in one pass, and shows the breakdown so you can sanity-check it against this article. For fitting known hardware — inserts, bolts, magnets, bearings by name — the Hole Tolerance Calculator has the specs library.