OpenAI releases ChatGPT Images 2.0 — AI image generator now writes multilingual text, creates infographics and maps with near-perfect accuracy

OpenAI unveiled ChatGPT Images 2.0 after weeks of testing on LM Arena (a third-party AI testing platform). The upgrade arrives just months after the company released GPT-Image-1.5 in December 2025, which itself improved instruction-following and visual quality.

The new model can generate images containing long blocks of readable text in multiple languages, realistic website screenshots and user interfaces, infographics, slide decks, maps, and manga — tasks that previously produced garbled or unusable text and layouts. Early testers reported the output quality felt 'flawless,' a significant leap from earlier versions that struggled with legible text placement.

For business professionals and students, this means ChatGPT can now replace design and graphics tools for quick presentations, reports, and visualizations — no need to switch between ChatGPT for writing and Figma or Canva for images. Content creators can generate manga storyboards or multilingual marketing materials without manual editing.

ChatGPT Images 2.0 is already available to select ChatGPT users; broader rollout timing and API availability for developers have not been announced, but the model tested as 'duct tape' on LM Arena indicates public release likely within weeks.