Humans still have the upper hand at the International Mathematical Olympiad (IMO), but AI is closing in fast. At this year's event, held in Australia, five young math whizzes notched perfect scores, while generative AI models from Google and OpenAI landed gold-level results by racking up 35 out of a possible 42 points—a first for the machines at this elite competition, and a score matched by about 10% of the 630 competitors, who were all under the age of 20. As a press release from Google explains, "Each country taking part is represented by six elite, pre-university mathematicians who compete to solve six exceptionally difficult problems in algebra, combinatorics, geometry, and number theory."
A not-yet-released version of Google's Gemini Deep Think solved five of the Olympiad problems within the 4.5-hour limit, a leap from last year when Google's AI only hit the silver medal mark by answering four questions correctly and needed days to finish. According to IMO president Gregor Dolinar, who said the two companies "privately tested" their models on the 2025 problems, Gemini's answers were "astonishing" in their clarity and precision—just not quite flawless. CBS News notes no information regarding how much computer power had been used was provided.
Axios flags one important aspect of the news: "Models that previously beat humans in Go, Poker and other competitions were trained specifically for those games. The new high-performing models are general purpose models."
story continues below
Still, contest organizers couldn't independently confirm just how much computer power or human help may have gone into the AI submissions. As AI's math prowess surges, experts see broader implications. Google's DeepMind CEO Demis Hassabis forecasts machines with human-level thinking abilities within a decade, citing an "exponential curve of improvement" fueled by growing investment and talent.
,
uBlock 
,
and
Ghostery 
)
