ChatGPT Images 2.0
- OpenAI launched ChatGPT Images 2.0, a major upgrade to ChatGPT's image-generation capabilities. - DALL‑E 2 and DALL‑E 3 will be retired on May 12, while GPT‑Image‑1.5 remains accessible via the API. - The model adds web access, multi-image support, and better text and background rendering, per OpenAI release notes and PCMag. ( )
OpenAI has replaced ChatGPT’s built-in image generator with ImageGen 2.0, a new model now live across all ChatGPT plans. (help.openai.com, openai.com) The rollout landed on April 21, 2026. OpenAI said ImageGen 2.0 is available to free and paid users, while “ImageGen 2.0 Thinking” is limited to paid users who select Thinking or Pro models. (help.openai.com, help.openai.com) ImageGen 2.0 adds three practical changes: it can use web search in thinking mode, generate multiple outputs in one request, and handle text inside images more accurately, including non-Latin scripts. PCMag reported users can generate up to eight pages at once, such as comic pages that keep characters and style consistent. (help.openai.com, pcmag.com) Image generators turn written instructions into pictures, but older systems often miss object placement, mangle words on signs and posters, or lose consistency across a series. OpenAI said the new model improves instruction following, text rendering, aspect-ratio handling, and visual reasoning. (openai.com, pcmag.com) That makes the update especially relevant for work that mixes layout and language, like menus, flyers, storyboards, ads, and social posts. OpenAI’s help center says ChatGPT Images can now add text, preserve transparent backgrounds, edit uploaded images, and generate in any aspect ratio on web, iOS, and Android. (help.openai.com) The launch also marks another step away from the DALL·E brand inside ChatGPT. OpenAI’s API deprecations page says DALL·E 2 and DALL·E 3 will be removed from the API on May 12, 2026, with GPT Image models listed as the replacement. (developers.openai.com) For developers, the older tools are not disappearing all at once. OpenAI’s image-generation docs still list GPT image models including gpt-image-1.5, and the platform guide says GPT image models support image generation and editing while DALL·E 2 keeps the older “variations” endpoint. (platform.openai.com, developers.openai.com) OpenAI has also left a fallback path for people who want the older look inside ChatGPT. Its help articles say ChatGPT users can still generate images with DALL·E through the separate DALL·E GPT, even as the default image experience shifts to ImageGen 2.0. (help.openai.com, help.openai.com) The immediate change for most users is simple: the image button in ChatGPT now points to a model built for text-heavy, multi-step image work, not just one-off pictures. The next deadline is May 12, when the DALL·E 2 and DALL·E 3 API shutdown is scheduled to take effect. (help.openai.com, developers.openai.com)