GPT Image 1 (GPT-4o image generation)

What it is

GPT Image 1 is OpenAI's image generation model, and it first arrived not as a standalone drawing system but as a native ability of GPT-4o. OpenAI released the feature in ChatGPT on 25 March 2025^[1][2], and it replaced DALL·E 3 as the image generator in ChatGPT^[3]. The same model was offered in the API under the name gpt-image-1^[5]. Two names, one model: image generation handled inside a general-purpose multimodal model, rather than by a separate system attached to a chat product.

Where it came from

In the lineage, GPT Image 1 is the successor of DALL·E 3, and was itself succeeded by gpt-image-2 in April 2026^[3][5]. The model reached the API as gpt-image-1 on 23 April 2025; OpenAI said at that point that more than 130 million users had created over 700 million images in the feature's first week^[7]. That figure is a measure of reach as much as of output: the image feature was not a specialist tool people sought out separately, but something already sitting inside a chat product used at very large scale.

Main characteristics

OpenAI described the model as natively multimodal, saying that it excels at accurately rendering text, at following prompts precisely, and at drawing on GPT-4o's knowledge base and chat context^[6]. The first two points address weaknesses that had long dogged image generators: lettering that comes out garbled, and prompts that are only loosely obeyed. Legible words matter for anything with labels, signs, or diagrams in it; tight prompt adherence matters whenever a user has a specific picture in mind. The third point is what "native" buys in practice — the image model can draw on what the surrounding conversation already contains, so a request does not have to restate everything from scratch.

Every image it generated carried C2PA metadata identifying it as coming from GPT-4o^[6]. C2PA is a metadata standard for recording where a piece of media came from, so an image produced this way carries a machine-readable record of the model behind it.

Where it sits

Inside ChatGPT it became the default image generator for Plus, Pro, Team and Free users, with Enterprise and Edu to follow, and it was also available in Sora^[6]. Follow-up versions came the same year: GPT Image 1 Mini on 6 October 2025 and GPT Image 1.5 on 16 December 2025^[4]. API lifetimes were scheduled in advance — gpt-image-1 was set to shut down in the API on 23 October 2026, and gpt-image-1-mini and gpt-image-1.5 on 1 December 2026^[5].

The model marks a change in how OpenAI presented image generation: rather than a separate line of image models under the DALL·E name, the capability is presented as a feature of the GPT-4o model itself, and the naming of the API version follows the general model family. Within about a year it went from a ChatGPT feature to an API model with two smaller and updated siblings, and then to a scheduled retirement in favour of a successor.