GPT-Image-2.5

text to imageimage editingreference images

Create and refine images with GPT Image 2.5. Turn a written brief or reference photo into product visuals, portraits, and illustrations, then fine-tune the details that matter.

developer

OpenAI

input

Text + Images

output

Images

price

From $0.036/imageLow quality at 1K on Flare

released

Sep 8, 2026

Loading workspace...

creative capabilities

More Control over the Image You Have in Mind

Text-to-Image Creation

Describe your subject, setting, and visual direction to develop a new image from a written brief.

Reference Photo Fidelity

Explore a different setting or style while keeping the person, pet, or product in your reference more recognizable.

Focused Image Editing

Request a specific change to a background, object, or line of text, with better preservation of surrounding details.

Consistent Revisions

Refine an image through successive edits with improved continuity, so earlier creative decisions are less likely to drift.

Detailed Visual Briefs

Bring composition, lettering, and style instructions together for posters, illustrations, and other structured visual concepts.

Flexible Output Quality

Choose from five explicit quality levels, or use auto, to balance image detail with generation time and cost.

compare models

GPT Image 2.5 Flare vs. Sunburst

GPT Image 2.5 Flare

Start with Flare for everyday image creation, social content, and rapid concept iteration. OpenAI positions it as the default API choice for most applications, combining improved image quality with faster generation than GPT Image 2.

everyday creation

GPT Image 2.5 Sunburst

Choose Sunburst when careful editing matters more than turnaround time. It is designed for demanding creative work, such as campaign artwork and polished product images, with greater precision and longer generation times.

precision editing

model details

GPT Image 2.5 Specifications

Official model capabilities for Flare and Sunburst. Available controls may vary by platform.

Text + Images

input types

Generation + Editing

image tasks

Flare / Sunburst

model variants

1024 × 1024 / 1536 × 1024 / 1024 × 1536

recommended sizes

low / medium / high / xhigh / max / auto

quality settings

PNG / JPEG / WebP

output formats

creative possibilities

What Can You Create with GPT Image 2.5?

A skincare bottle photographed on a studio backdrop, the kind of product image GPT Image 2.5 can generate from a reference photo

ecommerce

Product Images for Your Next Launch

Start with a product reference and explore a studio backdrop, seasonal setting, or new material concept. Specify which packaging details should remain unchanged, then inspect the result before using it in a listing.

marketing

Campaign Visuals Built around Your Message

Develop a launch poster or social graphic from a focused brief. Supply the exact headline, brand colors, and layout priorities, then revise individual elements as the campaign takes shape.

A studio portrait restyled with different lighting and wardrobe, the kind of edit GPT Image 2.5 can make from an authorized reference photo

portraits

New Looks for Familiar Faces

Explore a portrait in a different photographic style, outfit, or setting. Use an authorized reference photo and describe the facial features, pose, and expression you want to preserve.

An illustrated scene rendered in a consistent hand-drawn style, the kind of artwork GPT Image 2.5 can develop across a series

illustration

Visual Stories with a Distinctive Style

Turn an idea into an editorial illustration, storybook scene, or concept image. Define the medium, palette, and character details, then use the result as a reference for your next revision.

before you create

GPT Image 2.5 Notes and FAQ

Practical Creation Notes

Begin with the subject, intended use, layout, and style. Put any exact wording in quotation marks.

For image edits, state both what should change and what should stay. Review each revision before continuing.

Check lettering, faces, fine textures, and factual details before publishing. Generated images can still contain errors.

Choose a recommended size first. OpenAI supports custom dimensions, but resolutions above 2560 × 1440 are experimental.

Transparent output requires PNG or WebP and a supported transparency setting. JPEG cannot preserve transparency.

Check the selected model, available controls, and displayed price before generating. A provider announcement does not confirm availability on every platform.

Frequently Asked Questions

GPT Image 2.5 is OpenAI's image generation and editing model family, released on September 8, 2026. Its API variants are Flare and Sunburst. ChatGPT Images 2.5 is the related image creation experience in ChatGPT.

Flare is a practical starting point for frequent image generation and quick iterations. Sunburst suits projects that prioritize precise edits and refined visuals. Compare both on your own brief when the choice matters; Sunburst takes longer to generate.

Yes. Both models accept image inputs for editing. Provide a reference and describe the change you want. For iterative API workflows, preserve the relevant image or conversation context instead of assuming a new request remembers earlier edits.

Sketch, templates, on-image comments, and prompt sharing are ChatGPT interface features. They are not automatically included in a third-party interface or API integration. Use the tools available in your chosen application.

OpenAI API charges depend on input and image output tokens, with image size and quality affecting the total. On APIXO, check the selected model's displayed price and billing rules. A ChatGPT plan does not include APIXO usage.