OpenAI on December 16 released GPT Image 1.5, a new image generation model that becomes the default for creating and editing pictures in ChatGPT, and paired it with a redesigned Images section in the app. The company says the model produces images up to four times faster than its predecessor and that API prices for image input and output have fallen by about 20 percent.
The release caps a busy fortnight for OpenAI, which has shipped a series of models in quick succession after Google drew attention with Gemini 3 and its Nano Banana Pro image model in November. OpenAI did not frame the launch as a response to Google, but the timing and the emphasis on speed and editing quality put the two companies in direct competition for image generation.
What changed in the model
GPT Image 1.5 replaces the model OpenAI introduced in March 2025, which powered the viral wave of stylized images earlier in the year. According to OpenAI, the new model follows instructions more precisely, renders text inside images more reliably, and handles multi-step edits better: a user can ask for a series of changes and the model keeps the lighting, composition and existing elements consistent across them rather than losing track after a few rounds.
The older model has not been removed entirely. It remains reachable in ChatGPT, but only through a custom GPT, and GPT Image 1.5 is now the default route for anyone generating an image in a normal conversation. In the API, the model is available as gpt-image-1.5, and OpenAI says the lower per-image cost is intended to make high-volume applications more economical.
Ars Technica noted that the improved conversational editing also makes it easier to produce convincing altered photographs, a concern that has accompanied every step up in image model quality.
A dedicated place for images in ChatGPT
The more visible change for most users is the new Images tab in ChatGPT, which is rolling out to free and paid users. Instead of typing a detailed prompt, users can pick from preset visual styles such as pop art, sketch, plush toy and doodle, and apply them to uploaded photos or generated images. OpenAI’s goal, as described in its announcement, is to lower the barrier for people who do not know how to write prompts.
The tab also includes a discovery feed showing what other people are making, with examples such as holiday cards, background removal and portraits rendered in the style of famous paintings. The feed functions as a prompt library for users who want a starting point.
Remembering a face
OpenAI’s materials also describe a likeness feature that lets a user upload a photo of themselves once so that later image requests can include them without re-uploading. The company says this is rolling out gradually and it is not yet visible to all users. For people who repeatedly generate images of the same person or character, such as creators producing thumbnails or brand mascots, the feature removes a recurring step.
Taken together, the faster model, the lower API price and the consumer-friendly interface suggest OpenAI now treats image generation as a core product line rather than a demonstration feature. The competition with Google is likely to continue to drive quick iteration on both sides.
