Google's latest AI image generation model, Nano Banana Pro, has garnered significant attention for its advanced capabilities. It excels in following detailed instructions and creating complex scenes, utilizing tools like Google Search and Maps effectively. While it produces impressive infographics and can generate long text, it still shows uneven intelligence, particularly in areas requiring domain expertise. For instance, it struggles with technical tasks like designing electrical circuits, where human input remains superior. The model has transformed the creation of infographics, allowing users to generate visually appealing and factually accurate representations of data. Prompts that previously yielded mediocre results now produce high-quality outputs, marking a shift in the potential applications of AI in education and information dissemination. Users can create engaging visual narratives, such as illustrated maps for storytelling, enhancing the utility of AI beyond mere image generation. Despite its strengths, benchmarking for these new capabilities is lagging. Current methods for assessing AI performance in image generation, like manual checks and specific competitions, don't fully capture the model's effectiveness in practical applications. As Nano Banana Pro becomes more integrated into various fields, there will need to be better standards for evaluating its output. This model represents a significant leap forward, but the conversation around its limitations and the need for improved assessment methods is just beginning.