GPT Image 2 Review

GPT Image 2 is the image model OpenAI released on April 21, 2026, sold in ChatGPT as ChatGPT Images 2.0 and in the API as gpt-image-2. It is the first OpenAI image model that thinks before it draws, planning a layout, searching the web for reference and checking its own output. It returns up to eight images per prompt at up to 2K.

Text is the breakthrough reviewers keep returning to. Signs, menus, interface mockups and dense infographics come out spelt correctly, and non Latin scripts such as Japanese, Korean, Hindi and Bengali finally work. In public blind voting OpenAI's image models hold the top three places on the arena leaderboard, and the September 8 update, ChatGPT Images 2.5, made generation markedly faster, kept faces truer through edits and added a Sketch tool, for every ChatGPT user, free tier included.

The caveats are speed and cost. Complex images in reasoning mode can take minutes, the deeper reasoning is reserved for paid plans, and API output at 30 dollars per million image tokens adds up at volume. As an image model it writes no code, which explains the coding grade. For posters, menus and product mockups, it is the first model to reach for.