An artificial intelligence company has revealed that the AI chatbot ChatGPT attained perfect scores in nine subjects in this year’s unified college entrance examination in Japan. The research team from LifePrompt, a Japanese AI startup, conducted tests on models such as OpenAI’s GPT-5.2 Thinking and Google’s Gemini 3.0 Pro, using questions from the 2026 unified college entrance exam. These AI models were tasked with answering questions across 15 major subjects. The results indicated that OpenAI’s model achieved perfect scores in nine subjects, with an overall average score of 96.9, while Google’s model attained an average score of 91.4. The nine subjects where perfect scores were achieved included mathematics and chemistry. However, OpenAI’s model delivered its weakest performance in the Japanese language subject, achieving a 90% accuracy rate. This examination is a compulsory test for applicants seeking admission to Japan’s national universities and some private institutions, with around 500,000 candidates participating annually. The scores will be used as a reference for admissions at 813 universities and colleges. In a specific subject for the Human Sciences Category I major at the University of Tokyo, ChatGPT demonstrated an overall accuracy rate of 97%, significantly outperforming the levels seen in 2024 and 2025. It is reported that candidates achieving an 89% accuracy rate in this subject have a 50% chance of being admitted to the major. The head of LifePrompt noted that ChatGPT’s perfect score in mathematics was due to its enhanced information processing and text comprehension capabilities, although in the Japanese language test, it struggled to fully integrate textual information. Importantly, during the test, access to search engines was disabled, and all responses were generated independently using the model’s internal data.