Vintage large-format camera dissolving into pixels

Now available on LinkModel

GPT-Image-2.5 API

Sharper details. Faster generation. More precise edits.

OpenAI’s new image models bring sharper detail, faster generation, and more precise editing to production creative workflows.

Explore Flare Explore Sunburst

2 models

Flare for speed, Sunburst for precision

Up to 4K

Custom square, portrait, and landscape output

6 levels

Auto, low, medium, high, xhigh, and max

Alpha

Opaque or transparent backgrounds

Model selection

Flare for speed. Sunburst for control.

Both models generate and edit images. Choose the latency and precision profile that matches the production job.

Model	Best for	Profile	Access
01Flare	Most apps and high-volume generation	A fast, high-quality image generation and editing model with reference-based changes, custom dimensions, multiple quality settings, and transparent output for frequent visual production.	Try Flare API docs
02Sunburst	Premium creative and polished product imagery	A precision-focused visual generation and editing model supporting multiple references, mask-guided changes, custom dimensions, and transparent backgrounds for demanding creative production.	Try Sunburst API docs

Reference fidelity

Edit only what you ask for

Compare five reference-led edits at their original portrait or landscape ratio. Drag across each image to see what changed—and what stayed anchored to the source.

Held vintage photo of the ceramic artist in an indigo blazer

Held vintage photo of a ceramic artist in a terracotta work jacket

InputGPT-Image-2.5

Edit instruction

Inside the held print, replace only the terracotta work jacket with a deep-indigo tailored blazer.

Keep unchanged

Identity, pose, hands, print perspective, glare, wear, and both studio environments

Visual range

One model family, many visual languages

Move from cinematic world-building to surreal composition without losing control of lighting, depth, or the internal logic of the image.

Retrofuturist orbital city viewed through a monumental window — Retrofuturism
Cinematic world-building with controlled light, scale, and architectural detail.

Surreal science-fiction landscape folding into an inverted city — Sci-fi surrealism
Complex spatial concepts rendered as one coherent visual system.

Cyberpunk city reflected across space above a lone figure — Cyberpunk
Dense atmosphere, readable depth, and consistent environmental styling.

Production controls

From reference to finished asset

Build dependable creative workflows around fidelity, focused changes, multi-turn consistency, and production-ready formats.

Reference fidelity

Preserve the subject

Keep recognizable people, products, composition, and visual identity anchored to the source.

Focused changes

Change one element

Update a product, background, or line of copy while leaving surrounding details intact.

Iteration

Stay consistent

Carry prior decisions forward across multiple editing turns without rebuilding the asset.

Production output

Handle real layouts

Create legible text, transparent assets, UI concepts, diagrams, and campaign creative.

Production outputs

Built for assets people actually ship

Typography, repeated systems, presentation layouts, editorial illustration, packaging, and merchandise all belong in the same production workflow.

Mid-century poster series with bold geometric typography — Poster system

Impressionist cityscape with the Golden Gate Bridge — Editorial illustration

Ornate wedding invitation with structured typography — Invitation design

Vintage national park stamp collection — Collection design

Solar flare presentation with charts and explanatory copy — Presentation

Earth and galaxy rendered as a detailed blue mosaic — Material style

ChatGPT sticker packaging featuring a black cat — Merchandise

Prompt anatomy

Write prompts like creative briefs

Start with the artifact, then define the subject, visual direction, and constraints. For edits, make the preservation rules explicit.

Prompt

Inside the physical photo being held, replace only the ceramic artist’s terracotta work jacket with a deep-indigo tailored blazer. Preserve her identity and pose, the holder’s hands, the print’s perspective, glare, scratches, and both studio environments.

Deliverable

A precise wardrobe edit inside a handheld vintage print

Subject

Change the ceramic artist’s terracotta work jacket only

Direction

Deep-indigo tailoring integrated into the faded 1988 print

Constraints

Keep the holder’s hands, print wear, glare, identity, pose, and both studio environments unchanged

API quickstart

One endpoint, two quality profiles

Use the same image-generation endpoint and switch between Flare and Sunburst with the model identifier.

QuickstartImage generation

curl -X POST https://api.linkmodel.ai/v1/images/generations \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-image-2.5-flare",
    "prompt": "A campaign portrait with controlled studio light",
    "size": "2048x2048",
    "quality": "high",
    "background": "transparent"
  }'

FAQ

Frequently asked questions

What is the difference between GPT-Image-2.5 Flare and Sunburst?Flare is the default for most applications, with high-quality output at lower latency. Sunburst is intended for premium workflows that need tighter editing control and the highest image quality.Can GPT-Image-2.5 edit existing images?Yes. Both models accept text and image input and support focused edits that preserve subjects and unchanged details.What sizes and backgrounds are supported?The API supports common and custom dimensions up to 3,840 pixels per edge, including 4K output, plus automatic, opaque, or transparent backgrounds.Which model should I start with?Start with Flare for most production workloads. Use Sunburst when maximum quality or precision across complex edits is the primary requirement.