At the 2025 ICPC World Finals, Google (DeepMind) and OpenAI published results of controlled experiments in which their models demonstrated gold-medal-level performance. DeepMind reported Gemini 2.5 "Deep Think" solved 10/12 problems (gold-level), and OpenAI stated its reasoning models solved 12/12 in their tests. These runs were company-organized experiments, not official competition entries — ICPC medals were awarded to university teams. The announcements sparked debate about AI participation in competitive formats. (Sources: DeepMind blog 17 Sep 2025; OpenAI posts 17–18 Sep 2025; ICPC official results).
Google and OpenAI AI Models Achieve Gold-Medal-Level Performance at ICPC (Unofficial)
