Google Gemini 2.5 Becomes First AI to Win Gold in Programming Contest

Google DeepMind has announced a breakthrough in artificial intelligence that it compares to some of the biggest moments in AI history. Google revealed that its Gemini 2.5 model became the first AI to win a gold medal at an international programming competition. The event took place in Azerbaijan earlier this month.

The achievement is being called a “historic moment” in the journey toward artificial general intelligence (AGI). AGI refers to AI that can think and solve problems at a human level across different areas.

The challenge was no ordinary coding test. Gemini 2.5 solved a highly complex real-world problem that had stumped human programmers. The task involved moving liquid through a network of ducts and reservoirs. The AI had to calculate the fastest way to distribute it by analysing an infinite number of possibilities. Amazingly, it found the solution in less than 30 minutes.

Google Gemini 2.5 Becomes First AI to Win Gold in Programming Contest

Human teams from top universities in Russia, China, and Japan could not solve it. Out of 12 total tasks, Gemini 2.5 solved 10. Its overall score ranked second among 139 of the world’s best college-level programmers.

Quoc Le, vice-president of DeepMind, compared the achievement to IBM’s Deep Blue beating chess champion Garry Kasparov in 1997, and DeepMind’s AlphaGo defeating a human Go master in 2016. He called this new step even more important because the problem was closer to real-world science and engineering, not just a board game.

DeepMind said the model performed “as well as a top 20 coder in the world.” The company stressed that competitions like these require abstract reasoning, creativity, and the ability to design new solutions.

Experts believe this could lead to progress in areas like drug discovery and chip design, where abstract problem-solving is key.

See Also: Google Discover Feed Expands to Include Instagram and X in Bid to Keep Users Within Its Ecosystem

Not everyone agrees on the scale of the achievement. Stuart Russell, a computer science professor at UC Berkeley, warned against exaggerated claims. He said AI has already been good at programming tasks for years. But he admitted that solving competition problems correctly shows progress toward more reliable coding systems.

Michael Wooldridge from Oxford University said the achievement was impressive, but raised questions about how much computing power was used. Google only confirmed that it required more power than what is available in its $250-a-month AI Ultra service.

Despite doubts, some in the academic community see this as a turning point. Dr Bill Poucher, executive director of the International Collegiate Programming Contest, said Gemini’s success sets a new standard. He believes it marks the beginning of how AI tools will shape the next generation of problem-solving.

This achievement adds to a long list of AI milestones. From the invention of the Perceptron in 1957, to Deep Blue’s chess victory in 1997, to AlphaGo’s Go triumph in 2016, and AlphaFold’s breakthrough in protein science in 2020—each step has pushed AI forward. Now, Gemini 2.5 joins that list, showing that AI is no longer just about games. It is starting to solve problems that humans struggle with in the real world.