OpenAI and Google outdo the mathletes, but not each other

AI models of Openi And Google DeepMind Gelden-Medal scores in the International Math Olympiade (IMO) from 2025, one of the world’s oldest and most challenging mathematical competitions at high school level, the companies announced independently in recent days.

The results underline how fast AI systems are improving, and yet how evenly matches Google and OpenAi seem to be in the AI race. AI companies are fiercely competing for the public perception to move forward in the AI race: an immaterial struggle of “Vibes” that can have large implications for securing Top AI talent. Many AI researchers come from backgrounds in competitive mathematics, so benchmarks such as IMO mean more than others.

Last year, Google scored a silver medal on IMO The use of a “formal” system, which means that it required people to translate problems into a machine -readable size. This year, both OpenAI and Google have introduced ‘informal’ systems in the competition, which were able to take questions and to generate answers based in the natural language. Both companies claim that their AI models have correctly answered five of the six questions about IMO’s test, which scored higher than most high school students and the AI model from last year, without requiring a translation of human-machine.

In interviews with TechCrunch, researchers behind the IMO efforts of OpenAi and Google claimed that these Goudmedale performance represented through breaches around AI-reasoning models in unpretended domains. Although AI reasoning models tend to do well on questions with simple answers, such as simple math or coding tasks, these systems struggle on tasks with more ambiguous solutions, such as buying a great chair or helping with complex research.

However, Google raises questions about how OpenAi has carried out and announced its gold medal IMO performance. If you are going to participate in AI models in a mathematics competition for high school students, you can argue as well as teenagers.

Shortly after OpenAi announced the performance on Saturday morning, Google DeepMind’s CEO And researchers went to social media Slam OpenAI for the announcement of his gold medal prematurely – Shortly after IMO announced which high school students had won the competition on Friday evening – and because they have not officially evaluated the test of their model by IMO.

Thang Luong, a senior researcher from Google Deepmind and Lead for the IMO project, WAN told that Google was waiting to announce his IMO results to respect the students who participated in the competition.

Luong said that since last year, Google has been working with IMO’s organizers in preparation for the test and previously wanted the blessing and official assessment of the IMO president announcing his official results, what it did on Monday morning.

“The IMO organizers have their assessment guideline,” said Luong. “So any evaluation that is not based on that guideline cannot make a claim about gold medal level [performance]. “

Noam Brown, a Senior OpenAI researcher who worked on the IMO model, told WAN that IMO reached a few months ago with OpenAI about participation in a formal mathematical competition, but the chatgpt maker refused because it worked on natural language systems that it thought was worth striving. Brown says that OpenAi did not know that IMO carried out an informal test with Google.

OpenAI says that it has hired external evaluators-three former IMO medal winners who understood the assessment system to assess the performance of the AI model. After OpenAi heard from his Gold-Medal score, Brown said that the company reached contact with IMO, which the company then told to wait until after IMO’s Friday night prize sceremony.

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

Google is not necessarily wrong here has a more official, rigorous process to achieve its gold medal score but the debate can miss the larger whole: AI models from different leading AI laboratories improve quickly. Countries from all over the world sent their smartest students to compete with IMO this year, and only a few percent of them scored as well as OpenAI and Google’s AI models.

Although OpenAi used to have an important lead over industry, it certainly feels that the race is more closer to each other than a company would like to admit. It is expected that GPT-5 will release in the coming months and the company certainly hopes to give the impression that it still leads the AI industry.