ChatGPT Images 2.0

OpenAI rolled out ChatGPT Images 2.0 on April 21, adding a new image model that can follow longer instructions and edit images with more precision. (openai.com) The company said the model improves text rendering, supports more non-Latin scripts, and adds “advanced visual reasoning” for prompts that mix layout, objects, and written copy. OpenAI’s release notes say ChatGPT Images 2.0 is available across all ChatGPT plans. (openai.com, help.openai.com) For developers, OpenAI is exposing the model as `gpt-image-2` in the application programming interface and in Codex. OpenAI said the release is aimed at production work where images need to be readable, localized, and usable without heavy cleanup. (community.openai.com, openai.com) Image generators have long been good at style and weak at spelling. OpenAI is pitching this update as a fix for posters, diagrams, slides, comics, charts, and other formats where broken text has made generated images hard to use in real work. (community.openai.com, macrumors.com) OpenAI also tied the model more closely to ChatGPT’s broader workflow. In its system card, the company said a new “thinking mode” can use reasoning and live web search to plan an image, generate multiple images from one prompt, and turn a rough request into a more researched final output. (deploymentsafety.openai.com, help.openai.com) That pushes the product beyond one-shot image generation. OpenAI is framing Images 2.0 as a tool for end-to-end design tasks inside ChatGPT, where the same session can handle research, copy, layout, and revisions. (openai.com, deploymentsafety.openai.com) The technical changes are concrete. OpenAI’s developer announcement says `gpt-image-2` supports more aspect ratios and resolutions up to 2K, and its pricing page lists the new model as the company’s state-of-the-art image generation offering. (community.openai.com, openai.com) OpenAI is also updating the model’s knowledge base. The system card says ChatGPT Images 2.0 has “significantly enhanced world knowledge,” a change meant to help with prompts that depend on recognizable places, objects, symbols, and current visual context. (deploymentsafety.openai.com) The company paired the launch with a new safety document that warns more realistic image output can increase deepfake risks, including political and sexual imagery involving real people. OpenAI said it uses prompt filtering, image-input checks, and output blocking before results are shown to users. (deploymentsafety.openai.com) Four months ago, OpenAI had positioned GPT-Image-1.5 as the faster, more consistent successor to its earlier image tools. Images 2.0 is the next step in that same push: fewer garbled words, more faithful edits, and a tighter fit inside ChatGPT itself. (openai.com, openai.com)