ChatGPT Images 2.0 Launch

OpenAI released ChatGPT Images 2.0 on April 21, adding a new image model to ChatGPT and the application programming interface, or API. (openai.com) (community.openai.com) The company said the system can generate up to eight related images from one prompt, render outputs at up to 2K resolution, and support wider formats including 3:1 layouts for ads, slides, and product mockups. (community.openai.com) (venturebeat.com) Image models turn text instructions into pictures, but older systems often broke on small details like labels, charts, and repeated characters across a sequence. OpenAI said Images 2.0 improves text rendering, multilingual output, instruction following, and “visual reasoning,” which means using the prompt and image context together to decide what belongs in the frame. (openai.com) (petapixel.com) OpenAI positioned the release around work that needs readable, finished graphics instead of one-off art: diagrams, infographics, charts, posters, comics, and presentation slides. The company’s developer announcement said the model is built for “production workflows” where images need to be accurate, localized, and usable without heavy cleanup. (community.openai.com) (openai.com) The API rollout split the product into two tracks. OpenAI’s docs list `chatgpt-image-latest` as the default image model used inside ChatGPT, while developers can call `gpt-image-2` directly in the API and in Codex. (developers.openai.com) (community.openai.com) OpenAI also tied the launch to “thinking mode,” a workflow where reasoning models can research, transform inputs, generate variations, and check results before producing a final asset. That pushes image generation closer to document and design work, where the model is expected to lay out information, not just illustrate it. (community.openai.com) (macrumors.com) Pricing on the API model starts at $0.009 for a low-quality 1024×1024 image, $0.034 for medium quality, and $0.133 for high quality, according to OpenAI’s model page. The same page lists `chatgpt-image-latest` as available through the Responses, Chat Completions, and image generation endpoints, with no support for streaming or function calling. (developers.openai.com) Early reaction centered on whether the model can finally handle tiny readable text and consistent multi-panel outputs, two tasks that have tripped up image generators for years. OpenAI’s launch post and creator demos highlighted maps, menus, user interfaces, and storyboard-style sequences to show those gains. (openai.com) (youtube.com) The release lands four months after OpenAI’s December 2025 image update, which introduced GPT-Image-1.5 and a faster image experience in ChatGPT. Images 2.0 is the company’s next attempt to make generated visuals look less like rough drafts and more like finished files. (openai.com 1) (openai.com 2)