# GPT Image 2.5 Sunburst

> GPT Image 2.5 Sunburst is one of the two image models OpenAI released on 8 September 2026. Sunburst is the precision one: it holds fine detail, and it drifts less when you edit the same image several times in a row. That matters for posters, product shots and brand assets, where the third edit should not quietly redraw something you were happy with.

Canonical: https://routemux.com/openai/gpt-image-2.5-sunburst
Provider: OpenAI

## Key facts

| Field | Value |
| --- | --- |
| Model ID | openai/gpt-image-2.5-sunburst |
| Official model ID | gpt-image-2.5-sunburst |
| Input modalities | text, image |
| Output modalities | image |
| Capabilities | Image Generation, Image Edit |
| Upstream release | 2026-09-08 |
| Availability | available |

## Pricing

- RouteMux: $0.0186 per generation
- Official: $0.062 per generation
- Saving vs official: 70%
- Official price source: relay_per_call_2026_09, verified 2026-09-10

Billing is prepaid wallet, charged per successful request only.

## API access

Base URL: https://api.routemux.com

Supported protocols:

- openai_images

## What is GPT Image 2.5 Sunburst

GPT Image 2.5 Sunburst is one of the two image models OpenAI released on 8 September 2026. Sunburst is the precision one: it holds fine detail, and it drifts less when you edit the same image several times in a row. That matters for posters, product shots and brand assets, where the third edit should not quietly redraw something you were happy with.

It is slower per image than GPT Image 2.5 Flare, which is the trade you are making. Flare is the faster default; reach for Sunburst when the output is going somewhere that will be looked at closely. They are separate model ids, not two quality settings of one model, so you choose between them by name rather than by parameter.

Billing is per image, not per token. One image costs one fixed amount — however long your prompt is, however complex the scene. Budgeting is price times image count, with no test batch needed to find out what an image will cost.

Each call returns one image. The size parameter selects the aspect ratio rather than the resolution, and the dimensions that come back are the ones that count. The response follows the OpenAI Images shape, with the image itself returned as base64 in data[0].b64_json.

It speaks the OpenAI Images protocol, so /v1/images/generations and /v1/images/edits take the same request shape as calling OpenAI directly.

## Key capabilities

- **The precision tier** — Built for edit fidelity and fine detail rather than speed. Slower per image than Flare.
- **Steadier across edits** — Repeated edits to the same image are less likely to disturb the parts you did not ask to change.
- **Flat price per image** — One price per image, independent of prompt length or scene complexity. Budget is price times count.
- **Native OpenAI Images** — Same request shape as calling OpenAI directly — existing code mostly works unchanged.

## Other variants in this family

- GPT Image 2.5 Flare (`openai/gpt-image-2.5-flare`)
- GPT Image 2 (`openai/gpt-image-2`)

## Sources

- [OpenAI model docs](https://platform.openai.com/docs/models)
- [OpenAI pricing](https://openai.com/api/pricing/)

_Content reviewed by Jerry Fan on 2026-09-10._