As far as I can tell that only really affects the smaller models - GPT-4 / Claude / Gemini all seem pretty much impervious to weird whitespace in my experience.