ChatGPT Images 2.5: What’s new in OpenAI’s latest image model

OpenAI has rolled out ChatGPT Images 2.5, its latest image-generation model, with improvements aimed not just at making images look better but at giving users more control over how they create and edit them.

The company says users now generate more than 3 billion images every week across ChatGPT Images and its GPT-Image models in the API. Images 2.5 promises sharper details, more natural lighting and textures, better preservation of subjects in reference photos and more reliable editing across multiple turns.

It is also faster, with OpenAI claiming up to 50% lower image-generation latency compared with Images 2.0.


So, what is actually new?

Also Read: Meta launches AI agent Muse that can access other apps to send emails, make payments

Better at preserving your reference images

One of the biggest changes is how Images 2.5 handles reference photos.

If you upload a picture of a person or product and ask ChatGPT to put it in a different setting, composition or visual style, the model is now better at preserving distinctive features while changing the surrounding image.

This should mean fewer instances where a person looks noticeably different from the original photo.

OpenAI also says lighting and textures are more natural, making generated images more realistic.

More precise editing

Images 2.5 is designed to make only the changes you ask for, while leaving the rest of the image intact.

For example, if you ask it to change the background of a product photo, the model is better at preserving the product, composition and other elements rather than altering them unnecessarily.

This could be particularly useful for product photography, marketing creatives and branded content.

It remembers your edits

The model is also better at maintaining consistency across multiple rounds of editing.

So if you generate an image and then ask ChatGPT to change the background, modify an outfit and adjust the lighting, earlier changes are less likely to disappear or degrade as you continue editing.

In other words, users can refine an image over several prompts without having to repeatedly start from scratch.

Also Read: Anthropic researcher quits, says AI firms 'gambling with our lives'

Better at complex prompts

Images 2.5 is also better at understanding detailed visual instructions, according to OpenAI.

The model can handle more complex layouts, transparent backgrounds and real-world information, while also doing a better job of following specific visual styles.

This could help users create images with multiple elements, specific compositions or defined brand guidelines without the model drifting from the original brief.

You can sketch your idea

OpenAI is adding Sketch, which lets users draw directly in ChatGPT and use the drawing as a visual reference.

You could, for example, sketch the rough layout of a room, an outfit or an image composition and ask ChatGPT to turn it into a finished visual.

Users can access the feature by typing @Sketch in ChatGPT.

Templates and prompt sharing

OpenAI is also introducing templates for popular formats such as Posters and Merch, giving users a starting point instead of a blank prompt.

Another new feature lets users share the prompt used to create an image along with the image itself. Others can then use the same idea with their own photos and details.

What about developers?

OpenAI is releasing two new image models in its API.

GPT-Image-2.5 Flare brings the quality, editing and speed improvements to developers and is designed for most applications, including social content, product experiences and high-volume generation.

GPT-Image-2.5 Sunburst is aimed at more demanding creative workflows where greater precision is needed, such as campaign creatives and polished product imagery.

The bigger change

The biggest shift with Images 2.5 is that OpenAI is making image generation more iterative.

Instead of generating an image from a single prompt and accepting the result, users can start with a photo or sketch, make targeted edits and keep refining the same image while preserving earlier changes.

That makes Images 2.5 less like a tool that simply generates a picture and more like an AI image editor that can follow a creative brief across multiple steps.

Images 2.5 is rolling out to ChatGPT, ChatGPT Work and Codex users across desktop, mobile and web.