OpenAI and Google DeepMind independently announced that their AI models attained gold-medal scores in the 2025 International Math Olympiad (IMO), one of the world’s oldest and most challenging high school-level math competitions.
Despite the fact that Google and OpenAI appear to be evenly matched in the AI competition, the results emphasize the rapidity with which AI systems are advancing.

AI companies are engaged in an intense competition to establish a public perception of being advanced in the AI race.
This ethereal combat of “vibes” has significant implications for the recruitment of top AI talent.
Many AI researchers have a foundation in competitive mathematics, which is why benchmarks such as IMO are more significant than others.

Google achieved a silver medal at the IMO last year by employing a “formal” system, which necessitated humans to translate issues into a machine-readable format.

Both OpenAI and Google entered “informal” systems into the competition this year.

These systems were capable of ingesting queries and generating proof-based answers in natural language.

Both companies assert that their AI models accurately answered five out of six questions on IMO’s test, surpassing the performance of the majority of high school students and Google’s AI model from the previous year, without necessitating any human-machine translation.

In interviews with TechCrunch, the researchers responsible for OpenAI and Google’s IMO initiatives asserted that these gold-medal performances signal advancements in AI reasoning models in non-verifiable domains. AI reasoning models are generally successful in addressing questions that have clear answers, such as simple arithmetic or coding tasks.

However, they find it difficult to perform tasks that require more equivocal solutions, such as assisting with complex research or purchasing a high-quality chair.

Nevertheless, Google is raising concerns regarding the manner in which OpenAI
conducted and announced its gold-medal IMO performance. In the end, if you are going to submit AI models to a high school math contest, you might as well dispute like an adolescent.

Google DeepMind’s CEO and researchers took to social media to criticize OpenAI for prematurely announcing its gold medal, shortly after IMO announced the high schoolers who had won the competition on Friday night.
They also criticized OpenAI for not having their model’s test officially evaluated by IMO.

By the way, we refrained from making an announcement on Friday due to our adherence to the IMO Board’s initial directive that all AI laboratories disclose their findings only after the official results had been verified by independent experts and the students had received the recognition they merited.

— Demis Hassabis (@demishassabis) July 21, 2025
According to Thang Luong, a senior researcher at Google DeepMind and the project’s director, the company delayed the announcement of its IMO results in order to show respect for the students who participated in the competition. This information was reported by TechCrunch.
Luong stated that Google has been collaborating with the administrators of the IMO since last year to prepare for the test.
The company sought the IMO president’s approval and official evaluation prior to announcing its official results, which it did on Monday morning.

“The IMO organizers have their own grading guideline,” Luong stated. “Therefore, any assessment that is not founded on that guideline could not assert that it demonstrates gold-medal-level performance.”

Noam Brown, a senior OpenAI researcher who collaborated on the IMO model, informed TechCrunch that IMO contacted OpenAI a few months ago to inquire about participating in a formal math competition.
However, the ChatGPT-maker declined, as it was currently focused on natural language systems that it believed were more valuable. Brown claims that OpenAI was unaware that IMO was conducting an informal test with Google.

OpenAI asserts that it engaged third-party evaluators, including three former IMO medalists who were cognizant of the grading system, to evaluate the performance of its AI model.
Brown stated that OpenAI contacted IMO after learning of its gold-medal score. IMO advised the company to postpone the announcement until after IMO’s Friday night award ceremony.

IMO did not respond to TechCrunch’s request for comment.

Google is not inherently incorrect in this regard; it did undergo a more formal, rigorous procedure to obtain its gold-medal score.
However, the debate may overlook the more significant issue: AI models from numerous top-tier AI laboratories are advancing rapidly.
This year, the most talented students from various countries participated in the IMO competition, and only a small percentage of them achieved the same level of success as OpenAI and Google’s AI models.

Although OpenAI enjoyed a substantial advantage over the industry in the past, it appears that the competition is more evenly matched than any organization would like to acknowledge.
It is anticipated that OpenAI will release GPT-5 in the near future, and the company is eager to maintain its position as a leader in the AI industry.