I tested OpenAI's new Images 2.0 model against Gemini's Nano Banana Pro, and the results are closer than ever—but not equal.

The new model can generate structured, high-quality visuals, including comics, layouts, and multilingual designs, all directly from user prompts.