What infill percentage should I use for functional parts?

20%, and put the extra material into walls. That combination beats raising infill on almost every load case, and it is cheaper.

Here is the comparison on a 60 mm cube at 0.2 mm layers and 60 mm/s:

The four-wall option uses a third less plastic and finishes two hours sooner than the high-infill one — and carries bending loads better.

Why walls beat infill

Bending stress is highest at the surface of a part and near zero at its neutral axis, which for most geometries is the middle. Infill lives in the middle. Perimeters live on the surface.

Adding a perimeter puts solid material exactly where the stress is. Adding infill puts it where the stress is not, at greater cost in both plastic and time.

Where infill genuinely earns its place

• Compression through the thickness. Feet, pads, spacers and anything squeezed between two surfaces need interior material or the skins collapse into the void.
• Screw bosses and threaded inserts. These want a local solid region, which is a modelling decision rather than a global percentage.
• Anything printed in TPU. For a flexible material, infill density behaves as a hardness dial for the finished object, and the number becomes a functional specification.
• Very thin parts where there is no room for more perimeters, so infill is all there is.

Pattern matters less than people think

Gyroid, grid and cubic all perform similarly at the same percentage on typical loads. Gyroid is isotropic and prints without crossing itself, which makes it a reasonable default. Choosing a pattern is a small optimisation on top of a large one.

A working default

Four walls, 20% gyroid, 0.2 mm layers, printed in PETG rather than PLA if the part will see heat or impact. That covers the overwhelming majority of functional printing, and everything above is the reasoning behind it.

Model your own combination in the print time estimator, and see pattern detail in the infill patterns guide.