POST /api/v1/images/generations with model: gpt-image-2-5-flare for text to image, or add image_list to edit up to 16 reference images. Pick a resolution tier and a quality level, and read the finished image from the job by its img_uuid.GPT Image 2.5 Flare API for Batch Image Generation
GPT Image 2.5 Flare is the speed-tuned variant of OpenAI's GPT Image 2.5 for everyday batch work. Send a prompt to
Get API Key
Built for Volume, Tunable per Request
Draft low, finish high
quality accepts low (default), medium, high, xhigh and max; image_size accepts 1k (default), 2k and 4k. Iterate on prompts at 1k low for 0.32 token ($0.0064) per image, then re-run the chosen prompt at a higher tier.
Transparent cut-outs at no extra cost
Set background to transparent to get a PNG that keeps its alpha channel, ready for product listings, stickers and layered layouts. The price is the same as an opaque image.
Edit with up to 16 references
Pass image_list as [{"url": "https://…"}] with 1 to 16 images, each under 50 MB, and describe the change in prompt. With ratio: auto the output follows the reference image's shape.
Only model and prompt are required; everything else has a documented default.
GPT Image 2.5 Flare Request Fields and Limits
model and prompt (required)
<code>model</code> is <code>gpt-image-2-5-flare</code>. <code>prompt</code> takes 1 to 32,000 characters; keeping it under roughly 8,000 tokens gives the best results.
ratio
<code>auto</code> (default), <code>1:1</code>, <code>1:2</code>, <code>2:1</code>, <code>1:3</code>, <code>3:1</code>, <code>2:3</code>, <code>3:2</code>, <code>3:4</code>, <code>4:3</code>, <code>4:5</code>, <code>5:4</code>, <code>9:16</code>, <code>16:9</code>, <code>9:21</code> or <code>21:9</code>. <code>auto</code> follows the reference image when one is supplied.
image_size and quality
<code>image_size</code>: <code>1k</code> (default), <code>2k</code>, <code>4k</code>. <code>quality</code>: <code>low</code> (default), <code>medium</code>, <code>high</code>, <code>xhigh</code>, <code>max</code>. Together they decide the price.
background
<code>opaque</code> (default) or <code>transparent</code> for a PNG with an alpha channel. Does not change the price.
image_list
Optional reference images to edit, 1 to 16 objects of the form <code>{"url": "https://…"}</code>, each under 50 MB.
custom_callback_url
Optional public HTTPS URL. GoEnhance POSTs the job-detail body to it on each status change, retrying up to 3 times with a 3-second timeout.
Per image, set by quality and resolution tier. Reference images and transparency add nothing.
GPT Image 2.5 Flare API Pricing
low: from 0.32 token ($0.0064)
1k 0.32 token ($0.0064), 2k 0.64 token ($0.0128), 4k 1.07 tokens ($0.0214) per image. This is the default tier.
medium: from 0.71 token ($0.0142)
1k 0.71 token ($0.0142), 2k 1.45 tokens ($0.029), 4k 2.4 tokens ($0.048) per image.
high: from 2.84 tokens ($0.0568)
1k 2.84 tokens ($0.0568), 2k 5.78 tokens ($0.1156), 4k 9.61 tokens ($0.1922) per image.
xhigh: from 5.06 tokens ($0.1012)
1k 5.06 tokens ($0.1012), 2k 10.28 tokens ($0.2056), 4k 17.08 tokens ($0.3416) per image.
max: from 11.38 tokens ($0.2276)
1k 11.38 tokens ($0.2276), 2k 23.12 tokens ($0.4624), 4k 38.42 tokens ($0.7684) per image.
Charged on acceptance, refunded on failure
The submit response returns the charged <code>cost</code>. If the job ends in failure, the tokens are returned automatically.
Integrate the GPT Image 2.5 Flare API
1
1. Create an API key
Generate a key at /keys and send it as
Authorization: Bearer <key> from your server. 2
2. Choose the tier
Set
image_size and quality for the job; start at 1k and low while testing prompts, and add background: transparent for cut-outs. 3
3. Submit and store img_uuid
POST
{"model": "gpt-image-2-5-flare", "prompt": "…"} to /api/v1/images/generations, plus image_list when editing, and save the returned img_uuid. 4
4. Collect the image
Poll
GET /api/v1/jobs/detail or wait for your callback. On success, download the image from json[0].value within 24 hours, before link_expired_at.Workloads where many images are needed and the tier can be chosen per job.
What Developers Build with GPT Image 2.5 Flare
Catalog and listing imagery
Generate product shots and transparent cut-outs for marketplace listings, with low-tier drafts reviewed before higher-tier finals.
Ad and social variants
Produce the same concept across many aspect ratios, from 9:16 stories to 21:9 banners, by changing only <code>ratio</code>.
Reference-based edits
Restyle, recolor or recompose existing photos by sending them in <code>image_list</code> with an edit instruction.
In-app generation features
Offer image generation inside your own product and expose quality levels to users as price tiers.
Endpoint, tiers, editing and results.