ChatGPT Images 2.0

OpenAI has released ChatGPT Images 2.0, a new image model for ChatGPT that is built to make and edit more complex visuals than earlier versions. (venturebeat.com) Image models turn a written prompt into a picture, but older systems often broke on the hard parts: small text, charts, user-interface mockups, and multi-step edits. OpenAI said Images 2.0 is designed to hold onto those details and to make “precise edits” without changing the rest of the image. (techcrunch.com; openai.com) OpenAI’s earlier ChatGPT Images release arrived on December 16, 2025, with GPT Image 1.5 and a promise of faster generation and more reliable editing. The new 2.0 release follows about four months later and adds what OpenAI and reviewers describe as stronger prompt following, better text rendering, and more structured outputs such as slides, maps, floor plans, and image grids. (openai.com; venturebeat.com) One of the headline changes is scale inside a single request. VentureBeat reported that the model can generate up to eight distinct images from one prompt, which lets users ask for sets of ad sizes, comic panels, or character turnarounds instead of one image at a time. (venturebeat.com) Text inside images has been a weak spot for image generators for years because many systems learned to reconstruct pictures as a whole, not to spell letter by letter. TechCrunch found Images 2.0 could produce a restaurant menu that looked usable at a glance, and OpenAI said the model can handle small text, iconography, user-interface elements, and dense layouts at up to 2K resolution. (techcrunch.com; techcrunch.com) OpenAI told reporters the model has “thinking capabilities,” which TechCrunch said include checking its own work, searching the web, and planning multi-image outputs. In practice, that shifts the product from single-picture generation toward design tasks that look more like making a brochure, storyboard, or infographic. (techcrunch.com; venturebeat.com) The release also lands as OpenAI is still migrating users away from older image systems. OpenAI’s developer documentation still lists GPT Image 1.5 as its latest mainline image model and says DALL·E 2 and DALL·E 3 are deprecated, with support ending on May 12, 2026. (developers.openai.com) That means the company is now selling image generation less as a novelty and more as a general-purpose visual tool inside ChatGPT and its application programming interface. OpenAI’s Help Center says ChatGPT Images can create new images and edit existing ones directly in ChatGPT, and the API docs point developers to image generation and multi-turn editing workflows. (help.openai.com; developers.openai.com) The remaining gap is language coverage. OpenAI and outside reviewers say non-Latin scripts such as Japanese, Korean, Hindi, and Bengali are improved, but reports on early use still describe multilingual text as uneven compared with English-heavy outputs. (techcrunch.com; venturebeat.com) For OpenAI, the immediate test is whether people use Images 2.0 for everyday work — menus, mockups, slides, maps, and edits — rather than one-off art prompts. The company made the same bet in March 2025 with 4o image generation and again in December 2025 with GPT Image 1.5; this release pushes that utility pitch further. (openai.com; openai.com)