I think there’s a flaw in this benchmark.
You’re supposed to measure the time after the first loop because everything is measured at the end now, when all the time is added up together if I’m not mistaken.
If vector.magnitude() is faster however then I’m pretty surprised as to why that is.