What GPT-Image-2 actually changed
OpenAI shipped an image model this week that plans before it draws, searches the web while it composes, and checks its own work before returning the result. GPT-Image-2 scored 1,512 on Image Arena, 242 points above the next closest model, four times the largest jump anyone had recorded.
TL;DR
- GPT-Image-2 plans before drawing, searches the web while composing, and verifies its work.
- It achieved a score of 1,512 on Image Arena, significantly surpassing other models.
- The model's architecture allows it to reason through composition similarly to how text models reason through arguments.
- New workflows are now viable, including localized campaigns and coherent design systems from a single prompt.
- The article details the model's internal mechanisms, new workflows, potential for forgery, and implications for various professional roles.
- It also provides prompts for building a creative operations function.