Still, I guess if you would care you could let the LLM perform loops with pangram tests, changing the text till it scores well enough.