Google DeepMind’s Gemini 2.5 Solves Complex ICPC Problem, Claims Gold-Level at Azerbaijan Contest
Google DeepMind says its Gemini 2.5 model cracked a complex real-world ICPC problem that stumped human teams, earning a gold-level result at an international programming contest in Azerbaijan. The model solved most tasks in under 30 minutes and ranked second overall out of 139 competitors, though it failed two of 12 problems. Experts praised the technical leap but urged caution over hype, compute needs and real‑world generalisability.