This article explores an analysis of how major AI models performed on high school science exams. Testers had models like Doubao process English data and reorganize the results, showing Gemini’s absolute lead in science testing, with GPT-5.1 in second place, while Qwen-3 lagged behind. The article also discusses the capability differences among AI models, such as language style and image understanding (relying on image tools), and speculates that GPT-5.2 may make breakthroughs in certain subjects in the future. The content provides deep insights into the performance of AI models in academic testing, helping to understand the current technical level and limitations of large AI models.
