From Perfect Gaokao Math Score to AI-Assisted College Applications: How Far Have China's LLMs Come?

As the 2026 Gaokao concludes, a new narrative is emerging where AI is tested alongside human students. Chinese media Zhendongxi conducted comprehensive tests of large language models on the 2026 exam with remarkable results. Tencent's Yuanbao AI assistant achieved a perfect 150 of 150 on the 2026 National Standard Math Paper I and II, while scoring 142 in Chinese and near-perfect marks in English. The results signal that Chinese LLMs are approaching top-tier human performance on standardized academic tasks. In the Chinese exam, Yuanbao demonstrated sophisticated essay-writing. One essay titled "That Word Called Ordinary, I Finally Learned to Read It," was praised by former Gaokao essay grading team leader Shen Yi as "a genuinely moving piece" with potential for near-perfect marks. The essay showed self-reflection about understanding ordinariness, transitioning from youthful rejection to mature acceptance. In mathematics, Yuanbao employed test-taking strategies familiar to top human students, substituting answer options to verify solutions efficiently — a mature exam strategy alongside computational power. Beyond testing, the article explores how AI agents now assist in college applications. These agents integrate massive university datasets, understand personal preferences, and provide personalized decision-making recommendations. This leap from information provision to decision support marks an inflection point. The broader implication: Chinese LLMs are not only approaching global first-tier capabilities in exams but are beginning to assist in real-world, high-stakes personal decisions, transitioning from solving problems to helping people make life choices.