Finally officially announced ChatGPT Images 2.0 and its underlying API model gpt-image-2. This version supports image output up to 2K resolution, as well as adjustable aspect ratios ranging from 3:1 to 1:3 for wide and tall images. The core update this time is the introduction of a 'thinking' mechanism, allowing the model to perform online searches, multi-scenario generation tests, and self-correction before creating images. The knowledge base for the new version model is set to December 2025. According to the official test data, this iteration resolves issues in earlier large image models related to rendering small text, UI icon layouts, and generating content in non-English languages. Currently, the image generation service with thinking capabilities is available to Plus, Pro, and Business tier users, with a new feature added for generating scannable QR codes. PS: The barcodes on the back of books created in the video can actually be scanned—insane!!!