Overview

• Google DeepMind said a tailored Gemini 2.5 solved 10 of 12 finals problems at the ICPC in Baku, achieving a gold-level result outside the official ranking under organizer supervision.
• Chinese outlet 36KR, cited by multiple reports, said an OpenAI ChatGPT variant solved all 12 tasks, a claim that has not been independently verified.
• DeepMind reported that Gemini alone solved a subtask on optimizing fluid distribution through pipes and described a dynamic-programming method that applied a min–max theorem.
• Coverage notes both labs used specialized, higher-compute configurations not available to consumers, with reports indicating OpenAI combined ChatGPT 5 with an experimental model.
• Berkeley professor Stuart Russell praised the technical progress but said proclamations of epochal importance appear exaggerated.