one year on
OpenAI and Google DeepMind Achieve Gold-Medal Level at ICPC World Finals in Baku
Both companies’ models solve the same rigorous 12-problem set as 139 university teams, with OpenAI posting a perfect score and DeepMind cracking a problem no human solved.
Originally published Published on this site
BAKU, Azerbaijan — OpenAI and Google DeepMind have both achieved gold-medal-level results at the 2025 ICPC World Finals, the premier university programming championship held here on September 4. The companies announced their separate performances today, with OpenAI claiming its reasoning models solved all 12 problems under the same five-hour time limit as the 139 competing student teams — a score it says would have placed first among humans. DeepMind’s Gemini 2.5 Deep Think solved 10 of 12 problems, earning a gold medal — and cracking one problem, Problem C, that no human team in the contest solved.
Both AI systems competed live under official ICPC rules, receiving identical problem sets and test-case constraints. OpenAI’s GPT-5 solved 11 problems on its first attempt; an experimental reasoning model solved the 12th after nine submissions. DeepMind’s model solved eight problems within 45 minutes, finishing 10 total in 677 minutes — a time that would rank second overall. The news comes just two months after both companies achieved gold-medal scores at the International Mathematical Olympiad in July.
The ICPC, often called the ‘Olympics of programming,’ tests algorithmic problem-solving across graph theory, optimization, and other advanced topics. Final rankings award points only for perfect solutions under tight time pressure. Google DeepMind said that if the best AI and human solutions in the competition were combined, all 12 problems would have been solved completely and correctly.
The record
ICPC's global executive director called the result "a key moment in defining the AI tools and academic standards needed for the next generation."
The OpenAI researcher said "the next frontier is the discovery of new knowledge, which is the true milestone at the end of the day."
One year later — open only if you can handle spoilers
The sweep capped a 2025 run of competition milestones — gold at the International Mathematical Olympiad that July, now the premier collegiate coding contest — and hardened the view that frontier models had passed elite human programmers on well-posed problems. The caveat, often lost in the headlines, was that these results came from internal, compute-heavy configurations rather than the shipping chatbots. Over the following year the contest itself stayed human, while "agentic" coding became the industry's central battleground.
Filed
Compiled and edited by Adam Allcock.
The Weekly Replay · free by email
This week, one year ago — every Sunday.
One email each Sunday: the week's replayed AI news, with the one-year-later annotations included. Written like it's breaking — dated like it isn't.
Free · double opt-in · unsubscribe anytime · privacy