A paid Claude model answered every English reading and mathematics question correctly in a reporter-run test of the College Scholastic Ability Test (CSAT), completing in minutes a set of problems that students spend many years preparing to solve. Even the free versions of artificial intelligence (AI) tools delivered impressive results. Perplexity answered 71 of the 74 questions correctly, making just one clear error and leaving two graphic-based questions unanswered. ChatGPT got every English reading question right, but fell short in mathematics, answering 32 of 46 questions correctly, with two wrong answers and 12 left unconfirmed. The models were given the 2027 academic-year university entrance exam papers, and their responses were checked against answer keys published by the state-run Korea Institute for Curriculum and Evaluation. The test covered all 28 English reading questions, excluding the listening part, and 46 math questions across the common section and three electives. Claude’s Opus 5.5 model, with its effort level set to “High,” correctly answered all 74 English and m