GPT-Image 2.5 Flare API
AI ImageGPT Image 2.5 Flare — Text-to-image generation and reference editing, priced per image in 1K, 2K and 4K tiers. Quality does not change the price.
- Per successful call$0.025–$0.055
- OperationsText to image
- Image edit
- View tool pricing and billing rules
Your result will appear here once you run the model.
01
Where it fits
GPT Image 2.5 Flare — Text-to-image generation and reference editing, priced per image in 1K, 2K and 4K tiers. Quality does not change the price.
Make Flare the default 2.5 route: at the same size and quality it measured identical to Sunburst, so pick it whenever you want the 2.5 generation.
02
What it does
Text to image
gpt-image-2.5-flare
Turn a text prompt into a new image.
Image edit
gpt-image-2.5-flare
Change an existing image from a prompt and reference input.
03
In depth
GPT-Image 2.5 Flare is OpenAI's image generation and editing model on OmniaKey, one of the two ids that make up the 2.5 generation next to GPT-Image 2.5 Sunburst. Both take a text prompt or a prompt plus a reference image, and the catalog lists the same per-call tiers and the same two capabilities for each: text-to-image and image-edit. OpenAI places Flare as the everyday generation choice and Sunburst as the one to use when an edit has to be precise.
Flare earns its place on photographs of a single object: in our four-prompt run it produced the closest match to a product brief of the three models we compared, and it kept every listed item in a room edit. It is not the id for printed menus, where it drew a lighthouse and a birds sketch above the heading nobody asked for. The catalog lists the 2.5 ids one tier above GPT-Image 2 at every size, so keep bulk drafts on the older model and use Flare for the frames that ship.
Our runs left a mixed picture that is worth knowing before the first task. Flare made the best product photo in that comparison and spent the same image tokens as Sunburst, yet it added a coffee cup, a water glass and a lighthouse drawing to a menu prompt and miscounted objects in both jars. Against Seedream 5.0 Pro it matched GPT Image 2 in all six tests. Treat the first frame as a draft and look at the corners before you accept it.
Isolated product photography
Use Flare for one object on a built set, the brief where it did best in our tests: wet asphalt, neon reflections out of focus, droplets on the toe box. Write the frame as an exclusion list as well as a wish list, naming the shot type and forbidding people and props, then check the corners of the first frame.
Room edits
Change one element of an interior and keep the rest. In our living-room case Flare replaced a sofa with a green velvet one, added the plant we asked for and left the pictures, books, rug and spine lettering untouched. Send the source inline, list each item that must stay, and compare the result against the source side by side.
Client-facing variants
Because Flare and Sunburst sit on the same listed tier and behaved alike in our paired prompts, Flare can serve as the default 2.5 id for new work while a brief is still moving. Log the id and the parameters next to each accepted file, so a later request can be repeated on the same model.
04
Hands-on notes
In our four-prompt comparison, GPT Image 2.5 Flare made the best product photo of the three models and billed the same tokens as Sunburst.
Read the testOn our cafe-menu prompt, GPT Image 2.5 Flare drew an unrequested lighthouse and birds above the heading and added a coffee cup and water glass.
Read the testIn our counting test, GPT Image 2.5 Flare put six marbles in the left jar and four lemons in the right instead of the five and three we asked for.
Read the testIn our six-test run against Seedream 5.0 Pro, GPT Image 2.5 Flare matched GPT Image 2 in every test at the same 31-second median.
Read the testWhen we probed GPT-Image 2.5 Flare on 25 October 2026, the max quality setting produced the same image-token count as medium.
05
Limits to plan around
- A transparent background did not survive our 25 October probe: the request came back as an opaque RGB PNG, and asking for jpeg or webp also returned PNG, so a cut-out step belongs in the pipeline.
- OpenAI lists Batch for GPT-Image 2 but not for either 2.5 id, so an overnight job that depends on batch handling has to stay on the older model.
- Flare added props nobody asked for in our menu run, including a lighthouse drawing, a coffee cup and a water glass; keep prompts to a short whitelist and inspect the whole frame.
- Edits carry the source image inline inside a request capped at 1 MiB, so a camera original needs resizing or compression before it can be sent.
06
Compare nearby models
| Model | Best for | Operations | Listed price |
|---|---|---|---|
| GPT-Image 2.5 FlareThis page gpt-image-2.5-flare | Product photography and new 2.5 work, where our product brief was matched most closely. | Text to image · Image edit | $0.025–$0.055 per call |
| GPT-Image 2.5 Sunburst gpt-image-2.5-sunburst | Menus, price lists and other text-heavy layouts, where its spelling and alignment held in both runs. | Text to image · Image edit | $0.025–$0.055 per call |
| GPT-Image 2 gpt-image-2 | High-volume drafts and batch jobs, on the older id with the lower listed tier and explicit pixel sizes. | Text to image · Image edit | $0.02–$0.05 per call |
07
Sources and verification
Guide updated 2026-10-27.
08
Call it
Media models run asynchronously. Create a task, poll it until it finishes, then download the signed asset URL.
# 1. Create the job. 202 Accepted; the body carries the task id,
# the pre-charged price, and a poll_url. Reuse one Idempotency-Key
# across retries so a timeout never starts a second paid job.
curl --fail-with-body https://api.omniakey.com/v1/media/tasks \
-H "Authorization: Bearer $OMNIAKEY_API_KEY" \
-H "Content-Type: application/json" \
-H "Idempotency-Key: gpt-image-2.5-flare-001" \
-d '{
"model": "gpt-image-2.5-flare",
"type": "image",
"operation": "generate",
"input": {"prompt": "a paper boat on a quiet lake at sunset"},
"parameters": {"size": "1024x1024", "quality": "medium"}
}'
# 2. Poll GET /v1/media/tasks/{id} until status is
# succeeded | failed | canceled. Progress carries percent and stage.
curl --fail-with-body https://api.omniakey.com/v1/media/tasks/TASK_ID \
-H "Authorization: Bearer $OMNIAKEY_API_KEY"
# 3. Download assets[0].url before expires_at, or poll again for a
# fresh signed URL. Failed and canceled jobs are refunded.
curl -sS -o out.png "$(jq -r '.assets[0].url' result.json)"09
Pricing
Image jobs are billed per successful call, not per token. Failed or canceled jobs are refunded automatically.
$0.025–$0.055per callSave 53%
- 1K
- $0.025per call
- 2K
- $0.035per call
- 4K
- $0.055per call
More OpenAI image models
Related articles
GPT Image 2.5 API: Generate and Edit Images in Python
One POST creates the task and one GET reads it back. Here is the request code, the sizes we ran on October 29, 2026, and one edit shown before and after.
Read articleGPT Image 2.5 vs Nano Banana 2.1: Same-Prompt Test and Price
Four prompts, three image models, one day: what each made and what it really cost.
Read articleSeedream 5.0 Pro vs GPT Image 2: Same-Prompt Test and Price
Six jobs, the same prompts, three models: what each made and what it cost.
Read articleGPT-Image 2.5 vs GPT-Image 2
On OmniaKey both generations bill per image by size tier, and quality does not change the price. The 2.5 ids add no working quality tier, and requests need an explicit WIDTHxHEIGHT.
Read articleFrequently asked questions
How do I call GPT-Image 2.5 Flare on OmniaKey?
POST to /v1/media/tasks with the model id, type "image", and operation "generate" or "edit". Reuse one Idempotency-Key across retries so a timeout never starts a second paid job.
What does GPT-Image 2.5 Flare cost?
GPT-Image 2.5 Flare bills $0.025–$0.055 per call. Creation pre-charges that amount from your prepaid balance, and failed or canceled jobs are refunded automatically.
Why does the API return a task id instead of the image?
Provider latency runs from seconds to about a minute, so OmniaKey normalizes every image route into one asynchronous lifecycle. Poll GET /v1/media/tasks/{id} until the status is succeeded, then read the signed asset URL.
Is GPT-Image 2.5 Flare better than GPT-Image 2?
On our prompts the difference was narrow: Flare matched GPT-Image 2 in all six paired tests and gave the better product photo in another run, while sitting one listed tier higher at every size. Choose Flare for a new project and a single hero frame, and keep GPT-Image 2 for volume.
How do I send a GPT-Image 2.5 Flare request?
Create a task with POST /v1/media/tasks, model gpt-image-2.5-flare, and an Idempotency-Key header, then poll GET /v1/media/tasks/{id} until the state reads succeeded and download assets[].url. Put size and quality in parameters; for an edit, set operation to edit and pass the source image inline.
Does GPT-Image 2.5 Flare support transparent backgrounds?
Not in our test. A request with background set to transparent returned an opaque RGB PNG on 25 October 2026, and jpeg or webp output requests also came back as PNG, which leaves output_compression without an effect. Remove the background in your own pipeline and keep the request to size and quality.
Which size should I request?
Send an explicit width and height. OpenAI allows custom values when both edges are multiples of 16, the ratio stays between 1:3 and 3:1, no edge goes above 3840 pixels and the total lands between 655,360 and 8,294,400 pixels; anything above 2560x1440 is marked experimental. Confirm the dimensions in the downloaded file.
Can I batch GPT-Image 2.5 Flare jobs?
There is no Batch option on the 2.5 ids: OpenAI's pages list Batch support for GPT-Image 2 only. Schedule many small tasks instead, keep one Idempotency-Key per logical image, and resume a task by id rather than resubmitting it after a local timeout.