ChatGPT 5.5 Pro performs PhD-level math research independently

fields medalist timothy gowers used openai's chatgpt 5.5 pro to tackle open problems in number theory, with the model producing complete, verified scientific papers in under two hours without any mathematical guidance. in one instance, the model solved a problem in 17 minutes by swapping a standard proof component for a more efficient variant, improving an exponential bound to a quadratic one. when challenged with a harder problem, the model escalated its reasoning over roughly 40 minutes to achieve a polynomial bound, a result mit researcher isaac rajagopal described as the sort of idea a human would be proud to find after weeks of work. the model autonomously formatted the results into latex preprints, marking a significant milestone in ai's capability to perform high-level, novel mathematical research.