C11

If you do the same thing within numpy, you get a different result.

Numpy uses C underneath, and the C11 standard states that nan should be treated like it isn't there.

which brings me to two new additions to Python 3.15, math.fmax and math.fmin, which are max and min functions that correspond to the C11 standard.

So fmax3 and nan gives three, and fmax nan and three also gives three, which to me is a little more sane.

And so no matter which math standard your code needs, Python now supports both of them within the standard library.