ChatGPT Images 2.0 features
- ChatGPT Images 2.0 (gpt‑image‑2) is being used for complex multimodal outputs like infographics and character sheets. - It offers reasoning pre‑generation, up to eight coherent images per prompt, readable tiny text, multilingual labels, and ~2K resolution. - The model is available in ChatGPT, Codex, and the API, enabling consistent visual outputs for designers and developers ( ).
OpenAI has rolled out ChatGPT Images 2.0, a new image system it says improves text rendering, multilingual output, and visual reasoning inside ChatGPT. (openai.com) The release page, published April 21, 2026, says the model is built for image generation tasks that need readable labels and more structured layouts, not just photorealistic scenes. OpenAI’s developer forum post the same day says the underlying model, gpt-image-2, is aimed at diagrams, infographics, charts, posters, comics, and other text-heavy assets. (openai.com, community.openai.com) That shift targets a long-running weakness in image generators: they could draw polished pictures, but they often mangled small words, labels, and page-like layouts. OpenAI says ChatGPT Images 2.0 can handle multilingual text better and follow instructions more reliably from prompt to final image. (openai.com, community.openai.com) OpenAI is also tying image generation more closely to its reasoning models. Its developer announcement says those models can use a “thinking mode” to research, transform inputs, generate variations, and self-check before producing context-aware assets. (community.openai.com) The company is pushing the tool across consumer and developer products at the same time. OpenAI says gpt-image-2 is available in the API and in Codex, while the Help Center says ChatGPT Images 2.0 is live on all ChatGPT tiers, with “images with thinking” available on Plus, Pro, and Business and “coming to Enterprise and Edu soon.” (community.openai.com, help.openai.com) For developers, the practical change is control. OpenAI’s image API lets users request between 1 and 10 images in one call, choose output formats including PNG, JPEG, and WebP, and set size and quality options, while the image-generation tool in the Responses API can generate or edit images inside a chat workflow. (developers.openai.com, developers.openai.com) OpenAI’s forum post says gpt-image-2 supports more aspect ratios and resolutions up to 2K, and its ChatGPT help page says users can now generate images in any aspect ratio and edit them by selecting part of an image or describing a change in chat. That gives the same model a role in both one-off image prompts and repeatable production pipelines. (community.openai.com, help.openai.com) OpenAI is also framing the release as a replacement path for older image models. Its prompting guide recommends upgrading to gpt-image-2 for customer-facing assets, photorealistic generation, editing-heavy workflows, brand-sensitive creative, and text-in-image work where better first-pass quality can cut reruns and manual cleanup. (developers.openai.com) The company’s pitch is straightforward: image models are moving from art demos toward documents, ads, product screens, and other visuals people actually ship. ChatGPT Images 2.0 is OpenAI’s latest attempt to make that work inside the same tools people already use to chat, code, and build. (openai.com, community.openai.com)
