OpenAI's AI solved all 12 problems at the top university coding contest; the best team solved 11
At the 2025 ICPC World Finals, Google's Gemini also solved 10 of 12, including one that none of the 139 university teams solved.
The ICPC World Finals is the top programming contest for university students. In Baku in September 2025, 139 teams had five hours to solve 12 problems.
Google DeepMind’s Gemini 2.5 Deep Think solved 10 of 12, eight of them in the first 45 minutes, a result that would have ranked second. It also solved Problem C, which no university team solved. It competed live under contest rules, starting 10 minutes after the human teams.
OpenAI solved all 12 with an ensemble (several models working together) of GPT-5 and an experimental reasoning model. It submitted to the official judge alongside the teams. GPT-5 alone solved 11. The hardest problem took nine attempts. The best human team solved 11.
What it means
On timed, well-defined problems, AI systems now match or beat the world’s best student programmers. The results were not part of the official ranking, and neither lab said how much computing power it used. Contest puzzles are also very different from long software projects.
Caveats
- AI entries were not part of the official ranking. ICPC confirmed Gemini’s solutions were accepted but “did not validate the system, processes, or underlying model.”
- OpenAI’s result was announced by OpenAI; neither lab disclosed compute.
- Contest problems are short and well specified, unlike long-running software projects.