(It's specifically trained on formalized math problems, unlike most LLM, so it's not an apple to apple comparison.)