Yeah, LLMs used to not be up to par for new Project Euler problems, but GPT-5 was able to do a few of the recent ones which I tried a few weeks ago.