OpenAI vs Google: AI and Math Skills

AI Achieves Top Scores in International Math Olympiad

AI models developed by OpenAI and Google DeepMind have both attained gold-medal scores in the 2025 International Math Olympiad (IMO). This prestigious competition represents one of the oldest and most demanding high school mathematics challenges globally, as independently reported by both companies.

Rapid AI Advancement and Competitive Landscape

These results clearly demonstrate the accelerating pace of development within AI systems. Simultaneously, they highlight the remarkably close competition between Google and OpenAI in the ongoing AI race.

AI companies are engaged in intense competition for public perception of leadership in the field. This intangible “battle of vibes” can significantly impact the ability to attract and retain top AI talent.

Given the prevalence of competitive mathematics backgrounds among AI researchers, benchmarks like the IMO hold particular significance.

From Formal to Informal Systems

Last year, Google achieved a silver medal at the IMO utilizing a “formal” system. This required human intervention to translate problems into a machine-readable format.

This year, both OpenAI and Google submitted “informal” systems to the competition. These systems were capable of directly processing questions and generating proof-based answers in natural language.

Both companies assert that their AI models correctly answered five out of six questions on the IMO test. This performance surpasses that of most high school students and exceeds Google’s own AI model from the previous year, all without the need for human-machine translation.

Breakthroughs in AI Reasoning

Researchers from both OpenAI and Google, in discussions with TechCrunch, suggest that these gold-medal performances signify advancements in AI reasoning models within domains where verification is not straightforward.

While AI reasoning models generally excel at tasks with definitive answers, such as basic mathematics or coding, they often struggle with more ambiguous challenges, like selecting optimal furniture or conducting complex research.

Dispute Over Announcement Timing

However, Google has raised concerns regarding the manner in which OpenAI publicized its gold-medal IMO performance.

Following OpenAI’s announcement on Saturday morning, Google DeepMind’s CEO and researchers utilized social media to criticize OpenAI for prematurely announcing its gold medal. This occurred shortly after the IMO officially announced the winning high school students on Friday night.

Furthermore, Google contends that OpenAI’s model’s performance had not been officially evaluated by the IMO.

Respect for Participants and Official Grading

Thang Luong, a senior researcher at Google DeepMind and leader of the IMO project, explained to TechCrunch that Google deliberately delayed its announcement to show respect for the students participating in the competition.

Luong stated that Google had been collaborating with IMO organizers since the previous year. They sought the IMO president’s approval and official grading before releasing their results, which they did on Monday morning.

“The IMO organizers have established grading criteria,” Luong emphasized. “Therefore, any evaluation not adhering to these guidelines cannot legitimately claim a gold-medal level of performance.”

OpenAI’s Perspective

Noam Brown, a senior OpenAI researcher involved in the IMO model’s development, informed TechCrunch that the IMO contacted OpenAI several months prior regarding participation in a formal competition.

However, OpenAI declined, as they were concentrating on natural language systems they considered more promising.

Brown indicated that OpenAI was unaware of the IMO’s concurrent informal test involving Google.

OpenAI asserts that they engaged third-party evaluators – three former IMO medalists familiar with the grading system – to assess their AI model’s performance.

After receiving confirmation of their gold-medal score, Brown stated that OpenAI contacted the IMO, which requested they postpone the announcement until after the Friday night awards ceremony.

IMO’s Response

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

The Bigger Picture: Rapid AI Improvement

While Google’s approach involved a more official and rigorous process to achieve its gold-medal score, the core takeaway remains: AI models from leading labs are improving at an impressive rate.

This year’s IMO attracted the world’s most talented students, and a small percentage of them achieved scores comparable to those of the AI models from OpenAI and Google.

A Closer Race

Although OpenAI previously held a considerable lead, the current landscape suggests a more competitive environment than any company publicly acknowledges.

With the anticipated release of GPT-5 in the coming months, OpenAI aims to reinforce its position as a leader in the AI industry.