High reproducibility and consistency of the subject
Accurately maintains the facial features, texture, and lighting of the reference image, enabling stable depictions with consistent characteristics across continuous generations.
A precision image generation model designed for professional production workflows. It delivers high-fidelity visuals with up to 4K resolution and advanced subject consistency.
Each model has its own Playground, pricing, and field docs. Filter by capability to open the one you need.

OpenAI
High-precision image generation model that achieves professional detail rendering and precision editing

OpenAI
A professional-grade image generation and editing model that achieves the ultimate in high-precision editing and subject consistency.
Model id and a short note on what each capability is for.
| Model | ID | Description |
|---|---|---|
| GPT Image 2.5 Sunburst Medium(text to image) | gpt-image-2.5-sunburst-medium-text-to-image | High-precision image generation model that achieves professional detail rendering and precision editing |
| GPT Image 2.5 Sunburst Medium(image to image) | gpt-image-2.5-sunburst-medium-image-to-image | A professional-grade image generation and editing model that achieves the ultimate in high-precision editing and subject consistency. |
Featuring precise detail control and high consistency, it caters to demanding professional creative production.
In the magical forest, there is an elf girl with butterfly wings. She is gently enveloped by the glow of fireflies and the moonlight.
Prompt
An elf girl in a dreamy forest, with translucent butterfly wings growing from her back. Her long hair floats like a waterfall, surrounded by glowing fireflies and magical particles. Moonlight shines through the treetops,…

Generated with GPT Image 2.5 Sunburst Medium

Generated with GPT Image 2.5 Sunburst XHigh

Generated with GPT Image 2.5 Sunburst Max
A simple and spacious courtyard with a snowy landscape. A woman in a quiet kimono blends in with the morning light.
Prompt
A simple Japanese Zen garden, a snowy morning, a woman in a black kimono sitting in seiza on a tatami mat, gazing at the dry landscape of the rock garden. Outside the window, snow-covered pines and distant mountains, wit…

Generated with GPT Image 2.5 Sunburst Medium

Generated with GPT Image 2.5 Sunburst XHigh

Generated with GPT Image 2.5 Sunburst Max
The end-of-days desert, with motorcyclists racing forward, raising clouds of sand. A blood-red sunset and desolate ruins.
Prompt
The desolate desert of the apocalyptic world, a lone motorcyclist, wearing a tattered leather jacket and gas mask, riding a heavily modified heavyweight motorcycle, raising a huge cloud of dust behind him. A blood-red su…

Generated with GPT Image 2.5 Sunburst Medium

Generated with GPT Image 2.5 Sunburst XHigh

Generated with GPT Image 2.5 Sunburst Max
Typical jobs this series is used for.
With up to 4K high resolution and realistic texture rendering, you can efficiently create high-quality visual assets for advertising and e-commerce websites.
Utilizing up to 5 reference images, you can create multiple compositions and scenes while preserving facial features and product details.
Precisely modify specific elements or replace backgrounds while preserving the overall composition and the texture of unspecified areas.
Check production limits and starting credit cost before you pick another model for the same job.
| Model | Multimodal reference inputs | Max resolution | Sousaku starting price |
|---|---|---|---|
| GPT Image 2.5 Sunburst Medium | Up to 5 images | Up to 4K | 8 Credits |
| GPT Image 2.5 Sunburst XHigh | Up to 5 images | Up to 4K | 12 Credits |
| GPT Image 2.5 Sunburst Max | Up to 5 images | Up to 4K | 20 Credits |
From sign-up to your first successful call. The same steps work for every model.
Sign up and verify so you can issue an API key and use Playground.
Open a card above to see that model's details, pricing, and field docs.
Confirm inputs and parameters in the form before you automate.
The field docs on the detail page list required parameters. Use the same names in create_task.
Call the generate API with your key, poll until the task succeeds, then download the result.
Track credit use on the model's pricing page and in the dashboard.
Other series on Sousaku. One API key, one task flow.