GPT-Image-1

GPT-Image-1 is OpenAI’s native multimodal image generation model, the technology behind the image generator built into ChatGPT and offered separately through the Images API since April 23, 2025. Unlike the DALL-E series, which used a diffusion process, GPT-Image-1 generates images autoregressively within the same architecture family as OpenAI’s language models, which is why it follows detailed instructions and renders legible text inside images noticeably better than its predecessors. It supports text-to-image generation and editing at three quality tiers and three resolutions, including square, portrait, and landscape formats, with per-image pricing that scales from roughly a cent to about a quarter of a dollar depending on quality and size. It competes directly with Google’s Imagen and Midjourney for prompt-following and text rendering, and it has been widely adopted by developers building design and marketing tools on top of the API. OpenAI has already moved on to GPT-Image-1.5 and newer versions, and has set an October 23, 2026 deprecation date for the original GPT-Image-1 endpoint.