Current LLM’s are based on multi-character tokens, which means they don’t know how to spell well. As a result, they are horrible at spelling games like this or, say, Hangman.