Google DeepMind has the history of artificial intelligence with the Gold Medal win in the world’s most powerful mathematics competition

Google DeepMind It announced on Monday that a advanced version of the Gemini Artificial Intelligence model has officially achieved Performance at the level of the gold medal in International Sports OlympicsThe solution of five of the six difficult problems is exceptionally difficult and gain recognition as the first artificial intelligence system to receive official grades at the gold level from competition organizers.

Victory develops the field of thinking about artificial intelligence and puts Google in the intensive battle between technology giants to build artificial intelligence from the next generation. More importantly, it shows that artificial intelligence can now deal with complex mathematical problems using a natural language understanding rather than asking specialized programming languages.

“The official results in-Gemini achieved the level of medicine in the international sports drum!” Dimis HassanGoogle DeepMind, CEO, written on the X Monday Monday social media platform. “A advanced version managed to solve 5 out of 6 problems. Inspection progress.”

Official results in-Gemini achieved the level of medicine at the International Sports Olympics! ? An advanced version managed to solve 5 out of 6 problems. Incredible progress – our huge congratulations to LMThang And the team! https://t.co/p9bxf7rvj

– Demis hasabis (@demishassabis) July 21, 2025

the International Sports Olympics,, Which has been held annually since 1959, is widely considered the most famous mathematics competition in the world for pre -university students. Each country sends the participation of six young mathematicians to the elite to compete in solving six difficult problems exceptionally extending over algebra, compact, engineering and numbers theory. Only about 8 % of human participants usually earn gold medals.

How do you think Gueini Gemini deep in Google DeepMind

The latest success of Google exceeds its performance for 2024, when the company combined combined Alphabet and Eternal measurement The regulations obtained the state of the silver medal by solving four of the six problems. This previous system requires human experts to translate natural language problems first into field programming languages and then explain the sports product of Amnesty International.

This year’s penetration came Gemini deep thinkingAugmented thinking system that uses what researchers call “Parallel thinking“Unlike traditional artificial intelligence models that follow one series of thinking, deep thinking simultaneously explores multiple solutions before reaching a final answer.

“Our model works from one side to a party in the natural language, which results in strict sports evidence directly from the descriptions of the official problem,” Explanation of Hasabis In a follow -up post on the social networking site X, with confirmation that the system has completed its work within the standard time limit for the competition 4.5 hours.

We have achieved the impressive result of this year using an advanced version of Gemini Deep Think (augmented thinking mode of complex problems). Our model works from one side to a party in the natural language, which results in strict sports evidence directly from the descriptions of the official problem –…

– Demis hasabis (@demishassabis) July 21, 2025

The model has achieved 35 out of 42 possible points, comfortably exceeding the doorstep of the gold medal. According to the head of the International Maritime Organization, Professor Dr. Gregor Dolinear, the solutions were.Amazing in many respects“He found it” clear, careful and easy to follow “by the competition students.

Openai faces a violent reaction to overcome the rules of official competition

This advertisement comes amid increasing tension in the industrial intelligence industry on competitive and transparent practices. Google DeepMind approach to release its results from the artificial intelligence community, especially in contrast to the treatment of the Openai competition with similar achievements.

“We did not announce on Friday because we respect the request of the original IMO board of directors that all artificial intelligence laboratories participate in their results only after verifying the official results by independent experts, and students receive the right to increase they deserve,” Al -Hasabi BooksIt appears to refer to the previous Openai’s announcement of its Olympics performance.

Refer to the work aside, we did not announce on Friday because we respected the request of the original IMO board of directors that all artificial intelligence laboratories participate in their results only after verifying the official results by independent experts and receiving students the right they deserve

– Demis hasabis (@demishassabis) July 21, 2025

Social media users rushed to observe discrimination. “Do you see? Openai ignored the IMO request. Shame. No separation. Directly don’t respect,” Books one user. “Google DeepMind has acted with integrity, in line with humanity.”

Criticism stems from Openai’s decision to announce the results of its sports Olympiad without participating in the official IMO evaluation process. Instead, Openai had a committee of former participants in the International Maritime Organization to perform artificial intelligence, an approach to which some see in society as lacking credibility.

“Openai is likely to be the worst company on this planet at the present time,” one of the critics wrote, while others suggested that the company needs to “take things seriously” and “Be more credible.”

Do you see?

Openai ignore IMO request. shame. No category. Directly, do not respect.

Google DeepMind has acted with integrity, in line with humanity.

TVthnuke pic.twitter.com/8laoak6xue

– nik (@ns123abc) July 21, 2025

Inside the training methods that mastery of sports Gemini

The success of Google DeepMind stems from new training technologies that go beyond traditional methods. The team used the advanced reinforcement learning methods designed to take advantage of multi -step thinking, problem solving, and theories implementation data. The model has also been provided with a group of high -quality sports solutions and received specific guidelines about approaching IMO problems.

Technical achievement impressed the artificial intelligence researchers who have noticed its broader effects. “Not only a mathematics solution … but understanding the problems described for language and applying abstract logic to new cases,” Ai Observer Books Elyss Wren. “This is not a memory by heart – this is a movement emerging in the movement.”

Ethan MackeProfessor at the Warton School, which studies artificial intelligence, stressed the importance of using a model for general purposes instead of specialized tools. “Increase the evidence for LLMS’s ability to generalize to solve new problems,” he wrote, highlighting how this differs from previous methods that require specialized sports programs.

It was not just Openai.

Google also used a general model for purposes to solve very difficult mathematics problems at the International Mathematics Olympics in a clear language. Last year they used specialized tools

Increase evidence of LLMS’s ability to generalize to solve new problems https://t.co/ve72FMX2B

Emollick July 21, 2025

The model showed particularly impressive thinking in one of the problems as many human competitors applied mathematical concepts at the graduate level. According to the researcher DeepMind Junhyuk Jung, Gemini “made a wonderful note and used the theory of primary number only to create a self -list guide”, to find a more elegant solution than many human participants.

What does the Google DeepMind victory for the AI racing mean $ 200 billion

This penetration comes in a critical moment in the industry of artificial intelligence, as companies are racing to show superior thinking capabilities. Success has immediate practical effects: Google plans to present a copy of this Form of deep thinking Mathematics scientists are available for testing before it was presented to Google Ai Ultra subscribers, who pay $ 250 a month to reach the most advanced AI models in the company.

The timing also highlights the intense competition between major artificial intelligence laboratories. While Google celebrated its systematic and formal approach, the controversy surrounding Openai reflects broader tensions on transparency and credibility in developing artificial intelligence.

This competitive dynamic extends beyond mere sports thinking. Recent weeks have witnessed many artificial intelligence companies that announce penetration capabilities, although all of them are not received positively. Illon Musk recently launched Xai Groc 4The company claimed to be “the smartest artificial intelligence in the world,” though The leaders of the leaders showed that it was excessive Behind Google and Openai models. In addition, Grok has criticized controversial features Sexual intelligence comrades And birth rings Control content.

The dawn of artificial intelligence that thinks like humans-with severe consequences in the real world

The Olympiad Sports Learner’s victory exceeds the rights of competitive pride. Gemini’s performance shows that artificial intelligence systems can now match human level of thinking in the complex tasks that require creativity, abstract thinking, and the ability to synthesize visions across multiple areas.

“This is a great progress on the outcome of last year’s penetration,” Note DeepMind In their artistic announcement. The progress of the request for official languages is fully working in the natural language indicates that artificial intelligence systems have become easier and accessible.

For companies, these development signals that may deal with artificial intelligence soon with complex analytical problems in various industries without the need for specialized programming or experience in the field. The ability to think through the complex challenges in the use of daily language can pressure the advanced analytical capabilities across organizations.

However, the questions about whether this thinking potential will be effectively translated into challenges in the real world. Sports Olympics provides well-specific problems with clear success criteria-which are far from mysterious multi-faceted decisions that define most commercial and scientific endeavors.

Google DeepMind plans to return to the next year’s competition.In search of perfect degree“The company believes that artificial intelligence systems combine the fluency of natural language and strict thinking” will become invaluable tools for mathematicians, scientists, engineers and researchers, which helps us to enhance human knowledge on the road to AGI. “

But perhaps the most amazing details of the same competition appeared: When you face the problem of the most difficult competition, Gemini started an incorrect hypothesis and has never been recovered. Only five human students have solved this problem correctly. In the end, even the artificial intelligence that holds the gold medal still has something to learn from adolescent mathematicians.

Daily visions about business use cases with VB daily

If you want to persuade your boss at work, you have covered VB Daily. We give you the internal journalistic precedence over what companies do with obstetric artificial intelligence, from organizational transformations to practical publishing operations, so that you can share visions of the maximum return on investment.

Read with us privacy policy

Thanks for subscribing. Check more VB news bulletins here.

An error occurred.