ChatGPT passes Japan top universities’ entrance exams with highest scores

Sign up now: Get insights on Asia's fast-moving developments

The company tested using OpenAI’s ChatGPT 5.2 Thinking model, feeding it the exam questions converted into image data.

The company tested using OpenAI’s ChatGPT 5.2 Thinking model, feeding it the exam questions converted into image data.

PHOTO: REUTERS

Google Preferred Source badge

ChatGPT scored the highest marks in the 2026 entrance exams of the University of Tokyo and Kyoto University, two of Japan’s top universities, surpassing those of the actual top scorers, an artificial intelligence venture said on April 27.

According to LifePrompt, the generative AI chatbot scored 50 points higher than the top test taker on the University of Tokyo’s most competitive Natural Sciences III medical track exam and received a perfect score in mathematics. The achievement follows the AI’s failure to pass all of the school’s entrance exams in 2024.

The company tested using OpenAI’s ChatGPT 5.2 Thinking model, having it take the two universities’ undergraduate entrance exams and feeding it the exam questions converted into image data.

Since the answers included essay responses, they were graded by teachers from major cram school Kawai Juku.

ChatGPT also answered the 2026 unified university entrance examinations, and the Tokyo-based venture totalled the scores. ChatGPT scored 452 points out of a possible 550 on the University of Tokyo’s Humanities and Social Sciences exam, and 503 points out of a possible 550 on the Natural Sciences exam.

Both scores surpassed the highest scores of successful applicants announced by the university at 434 points for Humanities and Social Sciences III and 453 points for Natural Sciences III. On the other hand, the AI scored 90 per cent on the English exam but only 25 per cent on essay-style questions in subjects like world history.

As for Kyoto University’s examinations, the AI scored 771 points in the Faculty of Law exam, exceeding the highest passing score of 734, and 1,176 points in the Faculty of Medicine exam, above the top scorer’s 1,098.

In 2024, LifePrompt used OpenAI’s latest model, ChatGPT 4, to have AI solve the University of Tokyo’s entrance exam, but it failed to reach the minimum passing score. In 2025, LifePrompt tested the newest model, o1, which successfully cleared the passing threshold for the first time.

“The AI’s capabilities have been well documented. Given the rapid pace of AI evolution, companies will need to adopt AI with an eye toward how business operations will look in 10 to 20 years,” said LifePrompt head Satoshi Endo.

Professor Satoshi Kurihara of Keio University, who is also the head of the Japanese Society for Artificial Intelligence, argued that humans and AI should not compete on the same playing field because AI excels at absorbing vast amounts of existing data. KYODO NEWS

See more on