When this year’s entrance exam questions from the University of Tokyo and Kyoto University were fed into the AI chatbot ChatGPT, the results were striking. According to an analysis released on April 27 by Tokyo-based AI startup LifePrompt, the system outscored every successful applicant—effectively earning the title of top-ranked admittee.
In the University of Tokyo’s most competitive program, the Faculty of Medicine, ChatGPT scored 50 points higher than the highest human scorer. It also achieved a perfect score in mathematics.
University of Tokyo’s Yasuda Auditorium in Tokyo. (File photo ©Sankei/Hiroshi Watanabe)
Rapid Improvement Over Two Years
In the 2024 University of Tokyo entrance exams, ChatGPT failed to qualify for any faculty. Just two years later, it now sits at the top of the admissions rankings.
LifePrompt used OpenAI’s ChatGPT-5.2 Thinking model to tackle the second-stage written exams for the first-term admissions at both universities.
At the University of Tokyo, out of a possible 550 points, ChatGPT scored 452 in the humanities track and 503 in the science track—above the highest scores among admitted students, which were 434 and 453 respectively. This is all the more impressive given that this year’s mathematics problems were widely considered difficult, yet the AI still managed a perfect score. It also earned around 90% in English.
The exam questions were scanned and fed into the AI as images. ChatGPT’s responses were then graded by instructors from Kawaijuku, one of Japan’s leading university prep schools.
RELATED:
(Read the article in Japanese.)
Author: The Sankei Shimbun
Continue Reading