Humans beat generative AI models made by Google and OpenAI at a top international mathematics competition, despite the programmes reaching gold-level scores for the first time. US ChatGPT maker OpenAI said that its experimental reasoning model had scored a gold-level 35 points on the test.