This has been true in my experience. Chat gpt is already an awesome math tool. Significantly better than I expected it to be for applied math problems. Not sure how they accomplished the improvements in math performance GPT-5 demonstrates significantly improved mathematical accuracy compared to GPT-4, showing a "quantum leap" in its ability to solve complex math problems. On the American Invitational Mathematics Examination (AIME), GPT-5 achieved a 94.6% score without tools, a substantial jump from GPT-4's ~46.4% score on similar problems. With its built-in "chain-of-thought" or "Thinking" mode, GPT-5's accuracy on competition-style math problems can reach nearly perfect scores of 99.6%, a massive boost over GPT-4's similar-level performance.