On April 21st, OpenAI officially released the GPT Image 2 model, which surpassed Google’s Nano Banana2 in authoritative text-to-image benchmarks to secure the top position worldwide. The model has undergone several months of iterative improvements, notably boosting its image quality, comprehension, and detail rendering capabilities. According to SuperCLUE assessments, GPT Image 2 excels in key areas such as Chinese character generation, realistic reproduction, and overall image quality. It achieved a remarkable score of 93.07 in Chinese character generation, effectively addressing the long-standing issue of garbled or inaccurate Chinese text produced by previous overseas models. Additionally, the model demonstrates strong performance in accurately reproducing complex scenes, interpreting detailed prompts and logical requirements, and generating professional-grade visuals, including scientific diagrams and poster designs.