Natural light & texture
Late-afternoon sun, cracked crust, linen weave and wood grain — materials render with believable detail.
OpenAI
Quality or speed with GPT Image 2.5
OpenAI’s image model in two flavors — Sunburst for quality, Flare for speed. Natural lighting, rich textures and reference-guided editing with up to 16 images.
GPT Image 2.5 Studio
flare & sunburst
Describe any image — natural light, rich textures, readable text — and GPT Image 2.5 renders it faithfully.
Click or drop images here
1–16 images (JPG/PNG/WEBP, max 30MB each). Refer to them by order in the prompt, e.g. “keep the saxophone from image 1 and change the lighting”.Up to 20,000 characters. Sunburst keeps a reference character through edits with more visible detail.
Estimated cost 10 gems
Your image will appear here
What is GPT Image 2.5?
OpenAI’s September 2026 image model — Flare for speed, Sunburst for quality, one shared request format.
2 versionsFlare & Sunburst
16 refsReference-guided editing
13 ratiosIncl. 27:16 at 1K
Up to 4KTransparent at 1K
Why creators use GPT Image 2.5
Natural lighting, richer textures and faithful text — at two speed points.
Correctly spelled text
Poster briefs with title, subtitle and footer text rendered exactly as quoted in the prompt.
Reference characters
Keep a character through an edit — Sunburst preserves more visible detail like keys and tone holes.
Transparent stickers
PNG cutouts with real alpha channels at 1K — no separate background removal.
Flare when speed matters
The smaller model finishes faster with GPT Image 2-level quality — ideal for drafts and batches.
Same price either way
Flare and Sunburst share one rate card and one request — switch by changing the model ID.
How to create an image
Pick a version and mode, configure the output and generate.
1. Choose version & mode
Sunburst for quality or Flare for speed; Text to Image or Image to Image with references.
2. Write the prompt
Describe the subject, lighting and materials — or the edit, naming references by order.
3. Configure the output
Aspect ratio, resolution (1K–4K) and background — transparency is an option at 1K.
4. Generate and download
The studio estimates your cost live. Preview the result inline and download the image.
GPT Image 2.5 FAQ
Answers to common questions about versions, inputs and pricing.
What is GPT Image 2.5?
GPT Image 2.5 is OpenAI’s image model released in September 2026, in two versions: Flare, the smaller model optimized for speed, and Sunburst, the base model optimized for quality. Each runs as a text-to-image and an image-to-image model.
What is the difference between Flare and Sunburst?
Flare is OpenAI’s speed-oriented variant with image quality comparable to GPT Image 2. Sunburst is the base model optimized for quality — on LMArena it scores higher for text-to-image and single-image editing. Both take exactly the same request and cost the same.
What is the difference between Text to Image and Image to Image?
Text to Image needs only a prompt. Image to Image also takes up to 16 reference images — refer to them by order, for example “keep the character from image 1 and change the background to image 2”.
What resolutions and aspect ratios are supported?
Resolutions are 1K, 2K and 4K. Aspect ratios: auto, 1:1, 3:2, 2:3, 4:3, 3:4, 16:9, 9:16 and 21:9 — plus 27:16, 16:27, 9:8 and 8:9, which are available at 1K only.
How does billing work?
A flat per-image rate by resolution: 10 gems for 1K, 20 for 2K and 40 for 4K. Reference images are free — you only pay for the output.
Can I get transparent backgrounds?
Yes at 1K — set background to transparent for stickers and cutouts. At 2K and 4K the output may not keep transparency.
How many images do I get per request?
One image per request. Prompts can be up to 20,000 characters; iterate with references for variations.
Ready to create your next visual?
Start with a prompt or a set of references and generate images with GPT Image 2.5.
Start creating with GPT Image 2.5