What is GPT Image 2.5?

GPT Image 2.5 is OpenAI's September 2026 image generation family on NightCafe. It launched as two PRO models:

• GPT Image 2.5 Flare — the faster everyday model, and still the best first choice for most prompts.
• GPT Image 2.5 Sunburst — the higher-precision model. Use it when you want tighter control, more detailed creative work, or more careful edits.

Both models share the same Quality and Resolution settings. Quality defaults to Low, which is about 1 credit at Medium Res.

GPT Image 2.6 launched on 9 November 2026 and now sits above 2.5 in the Studio image picker. 2.6 costs more: Low at Medium Res is 2.5 credits, and it is not unlimited. For everyday work, stay on 2.5 Flare at Quality Low.

Independent Artificial Analysis text-to-image arenas (checked 9 November 2026) rank GPT Image 2.6 Sunburst (max) first and Flare (max) second. GPT Image 2.5 Sunburst (max) is third and Flare (max) is fourth. Those arena scores are for the Max quality setting — the expensive end of NightCafe's Quality ladder. For everyday use, Flare on Quality Low is still the first pick at about 1 credit.

What it's good at

• Prompt fidelity and typography — follows detailed briefs and renders readable text.
• Prompt to Edit — add a start image and describe the change.
• Multi-image fusion — up to 16 start images. Each start image adds 1 credit.
• Consistent characters — keep a subject looking the same across creations.
• Quality ladder — Low, Medium, High, Extra High, and Max. Start on Low; step up only when you need more polish.
• Resolution — Medium Res (about 1K) and High Res (about 2K).

How to use it

In Studio, pick GPT Image 2.5 Flare or GPT Image 2.5 Sunburst from the model picker. Write a prompt, optionally add start images, choose Quality and Resolution, then Generate.

You generate one image at a time. Quality is available in Controls so you can change it without opening More Settings.

Write a clear brief: subject, action, setting, camera, and any exact text in quotation marks.

A square product photo for an online shop. A ceramic pour-over coffee dripper in matte speckled cream sits on a walnut board, with a linen cloth and a small bowl of coffee beans. Soft window light from the left, gentle shadow to the right. The kraft card in front reads "HOUSE BLEND". Photoreal, clean background, no extra props.

To edit an existing image, add it as a Start Image (or use Prompt to Edit) and describe only the change.

Change the dripper from cream ceramic to glossy dark green, and keep the walnut board, linen, beans, lighting, and camera position the same.

Quality and credits

Credits are per image and depend on Quality and Resolution. Start images add 1 credit each.

On AI Creator and above, GPT Image 2.5 Low is unlimited when the generation costs 2 credits or less — so Low at Medium Res and Low at High Res do not use credits for a standard create. Higher quality settings still cost credits. Adding start images can take the job over the unlimited cap.

Higher-cost settings typically need Fast Credits.

Flare vs Sunburst

• Flare — start here. Faster to iterate, still a strong model, and the best credit value in the GPT Image lineup.
• Sunburst — switch here for extra precision, especially edits where the rest of the frame must stay put.

Both use the same credit table. You are choosing look and precision, not a different price.

Good to know

• Both models are PRO.
• Quality defaults to Low. Resolution defaults to Medium Res.
• You can attach up to 16 start images. Each one adds 1 credit.
• Output is one image per generate.
• GPT Image 2.6 is the newer generation if you want it and will pay 2.5 credits at Low. GPT Image 2 is still available if you prefer that earlier generation.

See also GPT Image 2.6 on NightCafe, GPT Image on NightCafe, and the NightCafe AI Models Guide.