H3 Max
AI Image Model

GPT Image 2.5

Choose the GPT Image 2.5 path that fits your next image

Generate a new visual or direct an edit with references, then choose GPT Image 2.5 Flare for a faster creative pass or Sunburst for deliberate detail control across up to 16 reference images.

Creative workspace

GPT Image 2.5

24

What is GPT Image 2.5?

GPT Image 2.5 lets you create from text or steer an edit with up to 16 reference images. Choose Flare for a faster creative pass, or Sunburst when the edit needs more deliberate control over the requested details.

How the people who built the model describe it:

“The OpenAI API lets you generate and edit images from text prompts using GPT Image models, including our latest, gpt-image-2.”
OpenAI image generation guide

GPT Image 2.5 at a glance

Variants
Flare + Sunburst
References
Up to 16 images
Output
1K, 2K, or 4K

What you can make with the GPT Image 2.5 tool

These are the jobs people actually bring to GPT Image 2.5, and the settings each one needs.

01

Fast concept exploration

Use Flare to move from a clear prompt to several creative directions, then keep the strongest direction for a more deliberate edit.

02

Reference-led product edits

Add product, material, palette, or composition references and state which visible details must stay stable while the scene changes.

03

Campaign systems and variations

Carry an approved visual direction into a set of formats, seasonal variations, or supporting assets without rewriting the brief from scratch.

GPT Image 2.5

Flare or Sunburst: choose by the job

Both variants work from a written brief and optional reference images. The practical choice is whether the next step benefits more from fast creative exploration or from more deliberate control over a sensitive edit.

Two clear creative paths

01

Two clear creative paths

Choose Flare when you are exploring a concept quickly. Choose Sunburst when the next revision needs more deliberate attention to the details you specify.

Use up to 16 reference images

02

Use up to 16 reference images

Bring together subject, product, color, material, composition, and style references in one request, then tell the model the role of each image.

Generate or edit from one workspace

03

Generate or edit from one workspace

Start with a new prompt when no source exists, or begin with a reference when preserving parts of an existing visual matters more than starting over.

Plan the delivery size before the run

04

Plan the delivery size before the run

Use 1K for early direction, 2K for a more detailed review, and 4K only when the approved image needs a larger final asset.

GPT Image 2.5

Choose with context

Where GPT Image 2.5 is the right call, and where a different model does the job better.

Flare

A faster route for creative exploration and new directions.

Sunburst

A deliberate route for detail-sensitive reference-led edits.

Reference capacity

Both variants accept up to 16 image references per request.

Starting point

Use a prompt for a new image, or references when selected details should guide the result.

How to work with GPT Image 2.5

01

Choose the working path

Pick Flare for a quick creative pass or Sunburst for a detail-sensitive edit.

02

Add only useful references

Upload up to 16 images and describe what each one contributes instead of asking every image to control every detail.

03

Write the constraints explicitly

Separate the requested change from the things that must remain stable, such as product shape, palette, or composition.

04

Review at the delivery size

Start smaller for exploration, then inspect text, edges, repeated objects, and important details before requesting a larger output.

GPT Image 2.5 questions, answered

Start Creating With GPT Image 2.5

Set up the brief now, then continue in the shared Studio workspace.