AI from OpenAI and Google win gold at 2025 Math Olympiad

Just a heads up, if you buy something through our links, we may get a small share of the sale. It’s one of the ways we keep the lights on here. Click here for more.

AI models created by OpenAI and Google DeepMind have achieved gold-medal scores in the 2025 International Math Olympiad (IMO), one of the most prestigious and challenging math competitions for high school students worldwide. 

Both companies independently announced that their AI systems solved five out of six problems, outperforming most human contestants, a major step forward in AI’s ability to tackle complex, proof-based math problems.

This marks a leap from last year, when Google’s AI only earned a silver medal using a more rigid system that required humans to reformat the questions. 

This year, both companies entered “informal” systems, capable of reading natural language and writing out detailed, human-like solutions, a closer match to how students solve problems in real life.

The news highlights two things: how quickly AI reasoning capabilities are improving, and how tight the race is between OpenAI and Google to be seen as the AI industry leader. 

Public perception matters in this space—top researchers often have math competition backgrounds, and wins at IMO act as credibility boosters.

However, the announcement led to a mini-controversy. OpenAI revealed its results on Saturday, just hours after IMO had officially awarded medals to students on Friday night. (Via: TechCrunch)

Google DeepMind researchers criticized OpenAI for announcing before its results were reviewed by IMO officials. 

Google, which worked with IMO organizers in advance, waited until Monday to reveal its results after receiving the green light.

OpenAI says it used third-party IMO medalists to grade its results and believed it had waited appropriately after the IMO ceremony. 

Still, Google argues that only the IMO’s official grading process can validate such claims.

Despite the disagreement, the larger takeaway is clear: AI is now performing at elite human levels in high school math, a domain previously seen as a major benchmark for human intelligence. 

With OpenAI set to release GPT-5 soon, the AI race is heating up, and the margin between competitors is getting razor-thin.

Do you think AI achieving gold medal scores in math competitions shows we’re getting closer to artificial general intelligence? Or are these specialized achievements not indicative of broader AI capabilities? Tell us below in the comments, or reach us via our Twitter or Facebook.