images25

images25 is an independent image generation and editing studio built on OpenAI's public API, centered on running the GPT Image 2.5 image generation and editing model released by OpenAI on September 8, 2026. The model is offered in the API as two models: gpt-image-2.5-flare and gpt-image-2.5-sunburst. images25 runs both models at the same time, letting users draft with Flare and finish final products with Sunburst without switching tools.

Main capabilities

• Dual-model support: gpt-image-2.5-flare is the default model, with higher quality than GPT Image 2 and up to half the latency, suitable for drafts, variations, and most daily work; gpt-image-2.5-sunburst is suitable for edits that must land precisely, slower but more accurate on text, faces, and layout, at the same price per token as Flare.
• Six quality tiers: the materials mention six quality tiers, with tiers such as medium, high, and xhigh appearing in examples.
• Explicit sizes and batching: supports explicit size settings, and up to eight reference images can be uploaded per request.
• Pre-run cost visibility: shows the credit consumption of a request before running it.
• Real outputs and timing: the site displays real generation results and their time taken, and calculates the cost per image.
• Prompt library: a built-in prompt library, categorized by creation type, including product photo prompts, UI mockup prompts, comic and storyboard prompts, infographic prompts, and packaging and typography prompts.
• Export and sharing: can export up to 4K, keep history, or publish a share link carrying the prompt.

How to use

1. Write a prompt: describe the scene first, then write the details; put text that must appear in quotation marks.
2. Choose Flare or Sunburst: choose Flare for speed and iteration; choose Sunburst when text or edits must be precise.
3. Download or share: export up to 4K, keep history, or publish a link carrying the prompt.

Use cases The prompt library examples in the materials cover: a two-color risograph concert poster, an amber serum dropper bottle, a desktop analytics dashboard, a four-panel restaurant comic, a quarterly revenue bar chart, a letterpress wine label, a coastal trail map poster, a stoneware mug in morning light, and more. Suitable for product images combining rapid drafting and precise final images, UI Image creation such as models, comic storyboards, infographics, packaging layout, etc.

Limitations and notes

• New accounts receive 10 image credits, i.e. 10 1K medium-quality images, with no credit card required; afterward, credits come from plans or one-time credit packs.
• All paid plans and free credits include commercial use rights to generated content, and users are responsible for the prompts they write; generations rejected by the model's safety filter are not charged.
• Reference images are used only for that generation and are not made public; the generator reads reference images on the server side, rather than handing public links to the model provider.
• Generated images are stored so they appear in history and so share links remain valid; uploaded reference images are private to the account only; unless users create a share link themselves, generated content is not made public.
• images25 is an independent studio, model names and specifications come from OpenAI's official documentation, and each content page notes the date it was checked against the reference documentation.
• Differences from ChatGPT: ChatGPT provides a chat window, while images25 directly provides model settings, including two API models, six quality levels, explicit sizes, batching, and the credit cost before running, and keeps a library of prompts and their corresponding parameters.
• The materials do not provide specific plan prices, credit pack prices, the specific credit consumption for each quality level, the list of supported specific sizes, API calling methods, or platform client information.