Short answer: It depends. Git stores "objects" as zlib compressed data. So the exact size of a commit will depend on the compressibility of your files (which is basically a function of the amount of repetition in them).