AI models excel in the university entrance exam in Japan

Artificial intelligence models have demonstrated remarkable performance in the university entrance exam in Japan in 2026, with OpenAI taking first place, achieving top scores in nine out of the 15 subjects evaluated. The experiment, conducted by Nikkei and the startup LifePrompt, showed that OpenAI's GPT-5.2 Thinking achieved an overall score of 96.9 out of 100, while Google's Gemini 3.0 Pro recorded 91.4.

The exam includes 21 subjects from seven disciplines, and the average scores among this year's candidates for the most popular 15 subjects was estimated at 58.1. The analysis highlighted a rapid improvement in AI performance, from 66 points in 2024 to 91 in 2025. However, AI models faced difficulties with the Japanese language and geography, underscoring their limitations in recognizing complex graphic information.