According to quantum mechanics--and experiment--photons don't have a specific position or energy until you go to measure one of these properties. Each photon has a position distribution (a size, if you will) and an energy distribution (i.e., a frequency spectrum).
Let's say that you measure a photon's position. The more accurately you measure its position--and as far as we are able to measure, real photons can be localized to an arbitrarily small point--the greater becomes the distribution of its energy. The more accurately you measure its energy (frequency), the larger becomes the distribution of its position. The product of these uncertainties is greater than or equal to Planck's constant divided by 2 pi. This relationship is known as the Heisenberg Uncertainty Principle. I want to stress that this principle is a statement about the nature of the properties of the photon and not a consequence of our specific methods of measuring them.
So in answer to your question, a photon's "size", if you want to call it that, depends not on its frequency, but on its spread of frequencies.
[Geek alert: math nerds may understand this wording better: momentum is the Fourier transform of position.]
[Insufferable geek alert: when we go to measure the size of a real photon, the answer, so far as we are able to determine, is that it is pointlike. The story for virtual photons, however, is different. They not only have size, but shape! This is because the virtual photons pull quark-antiquark pairs out of the vacuum. These form little spacetime "loops" with which other virtual photons can interact. The variation of the photon structure function F2 with momentum-transfer-squared is one of the most important experimental tests of quantum chromodynamics (the theory of the strong nuclear force).]