What Is GPT Image 2.5 Flare? What an Image Actually Costs, Measured

TL;DR: GPT Image 2.5 Flare (gpt-image-2.5-flare) is OpenAI's image model for "fast, high-quality everyday image generation", with a default snapshot dated September 8, 2026. On our API key, a high-quality 1024x1024 image billed 1,756 output tokens, about $0.053 at OpenAI's list price with the prompt (measured September 15, 2026). On TulexAI it costs 63 PT per image, Pro plan and up.

What is GPT Image 2.5 Flare?

GPT Image 2.5 Flare is an OpenAI image model that takes text and images as input and returns an image. OpenAI's model page describes it in one line as "Fast, high-quality everyday image generation" and calls it "our fastest model for high-quality, everyday image generation". The default snapshot is gpt-image-2.5-flare-2026-09-08, so it is a September 2026 release. It has a sibling, GPT Image 2.5 Sunburst, covered further down.

Here is what OpenAI documents for it, from the model page and the image generation guide (both checked September 15, 2026):

Spec	GPT Image 2.5 Flare
Model ID / default snapshot	gpt-image-2.5-flare / gpt-image-2.5-flare-2026-09-08
Input / output	Text and image in, image out (no audio, no video)
Endpoints	Image generation (v1/images/generations) and image edits (v1/images/edits)
Quality settings	low, medium, high, xhigh, max, auto (default: auto)
Recommended sizes	1024x1024, 1536x1024, 1024x1536
Custom sizes	Multiples of 16, aspect ratio from 1:3 to 3:1, no edge above 3840px
Also documented	Reference images, masks, transparent backgrounds, several images per request, PNG/JPEG/WebP output, streaming partial images
Default tier-1 rate limits	100,000 tokens per minute, 5 images per minute
Token prices (per 1M)	$5.00 text in, $1.25 cached text in, $8.00 image in, $2.00 cached image in, $30.00 image out

What does a GPT Image 2.5 Flare image cost?

OpenAI prices Flare per token on its pricing page, publishes no per-image price table for it, and its GPT Image 2 cost calculator does not estimate 2.5 token use. So on September 15, 2026 we generated images and read the usage OpenAI reported. Output image tokens, and what they cost at $30 per million:

Size	Quality	Output image tokens	Output cost at list price
1024x1024 (square)	low	196	$0.0059
1024x1024 (square)	medium	439	$0.0132
1024x1024 (square)	high	1,756	$0.0527
1536x1024 (landscape)	low	158	$0.0047
1536x1024 (landscape)	medium	343	$0.0103
1536x1024 (landscape)	high	1,372	$0.0412

The math for the bold row: 1,756 tokens x $30 / 1,000,000 = $0.0527. Our short text prompt added about $0.0002 at $5 per million, so a high-quality square image came to about $0.053, and a high-quality landscape image to about $0.041. A medium square image was about $0.013.

Three things in that table are worth knowing:

• High bills exactly 4x the output tokens of medium at both sizes (1,756 vs 439 square, 1,372 vs 343 landscape).
• Landscape billed fewer output tokens than square at all three quality levels we measured.
• auto resolved to medium at both sizes, and repeat runs returned identical output image token counts.

What an attached reference image adds

Image input has its own rate: $8 per million tokens. In our edit test, one 1024x1024 reference image at medium quality billed 1,024 image input tokens ($0.0082) plus the same 439 output tokens as a medium generation, about $0.021 before the prompt. Input is billed per token, so more reference images and longer prompts add to that.

Flare vs Sunburst: which GPT Image 2.5 model should you pick?

OpenAI ships two 2.5 models and gives one sentence on choosing between them in its image generation guide: "Choose Sunburst for workflows where editing precision matters most, and Flare for fast, high-quality everyday image generation."

That is the guidance OpenAI's guide gives on choosing between the two, and we add no ranking on top of it. Of the two 2.5 models, TulexAI carries Flare only. If editing precision is what your workflow needs most, OpenAI points you to Sunburst, which TulexAI does not offer.

How does GPT Image 2.5 Flare compare to GPT Image 2?

Per-token prices are identical. What differs is how many output tokens each bills at the same quality name (1024x1024, same key and day):

Quality	GPT Image 2.5 Flare	GPT Image 2
low	196 tokens	196 tokens
medium	439 tokens	1,756 tokens
high	1,756 tokens (~$0.053)	7,024 tokens (~$0.211)

At medium and high, GPT Image 2 bills 4x Flare's output tokens. A Flare high-quality render bills exactly the output tokens of a GPT Image 2 medium render. That is also how the two sit on TulexAI: Flare runs at its high setting and GPT Image 2 at its medium setting, so a square image bills the same 1,756 output tokens on either, and both cost 63 PT. Token counts are billing, not a quality score: they tell you what OpenAI charges, not which picture you will prefer. Quality levels, custom sizes, transparent backgrounds and mask edits are documented for the 2.5 models in OpenAI's guide and not documented for GPT Image 2, whose model page lists no quality settings (our key did accept low, medium, high and auto for it). For the full side-by-side and which one to use for which job, see GPT Image 2 vs GPT Image 2.5 Flare.

How do you use GPT Image 2.5 Flare on TulexAI?

Flare is one of the image engines on the TulexAI AI image generator. The details:

• Price: 63 PT per image, the same as GPT Image 2 (which TulexAI runs at its medium setting).
• Plans: Pro ($23/mo) and up. The $11 Basic plan does not include Flare; its only image engine is GPT-5 Image Mini.
• Quality: every Flare render on TulexAI runs at OpenAI's high setting. You never land on auto, xhigh or max. A square TulexAI render matches the 1,756-token row above and a landscape one the 1,372-token row; we did not measure portrait.
• Sizes: square (1024x1024), landscape (1536x1024) and portrait (1024x1536). Any other size comes back square.
• Edits: attach a photo with your prompt and Flare edits it rather than drawing a new picture. The edit runs on Flare itself.
• Heavier edits cost more: an edit with several reference images or a very long prompt counts as more than one image, in proportion to the usage OpenAI reports for that request.
• Output: one PNG per request.

More prompt examples are on the GPT Image 2.5 Flare model page, plans on the pricing page, and the wider field of image tools in our Midjourney alternatives roundup.

What prompts work well with GPT Image 2.5 Flare?

TulexAI renders Flare at high quality every time, so the prompt is your main lever. Here are three more to start from, alongside the examples on the model page.

1. A flat lay with a counted object list (square)

Overhead flat lay on a pale oak table, square frame: an open notebook with a pencil across it, a white ceramic mug of black coffee in the top right corner, a pair of round reading glasses, and three dried orange slices. Soft daylight from the left, gentle shadows, no writing or logos on any object, nothing touching the edge of the frame.

Why it works: a counted, placed list ("three", "top right corner") is a checklist you can verify at a glance. "No writing or logos" asks the model not to invent labels, and keeping objects off the edge leaves room to crop.

2. An editorial illustration with a fixed palette (portrait)

Vertical editorial illustration for a magazine article about sleep: a woman asleep on a thin crescent moon that rests on a stack of three books, placed in the lower half of the frame. Flat shapes, visible paper texture, a palette of indigo, cream and one coral accent only.

Why it works: the style is described by its materials rather than an artist's name, and "only" caps the palette at three named colors. Choose the portrait size so the canvas matches the vertical composition.

3. A one-region edit on an attached room photo

Using the attached photo: change only the wall behind the sofa to deep olive green paint with a matte finish. The sofa, rug, lamp, window and daylight stay exactly as they are. Do not add any new objects.

Why it works: it names one region to change, lists what you want kept, and rules out decorative additions, which tells the model what to leave alone. Check the result against your original: OpenAI recommends Sunburst over Flare when editing precision matters most. Keep to one attached photo where you can: several reference images or a very long prompt count as more than one image.

How we measured

On September 15, 2026 we called gpt-image-2.5-flare with our own OpenAI API key, using the same short text prompt (33 tokens) for every generation at 1024x1024 and 1536x1024, at low, medium, high and auto. The edit test sent one 1024x1024 reference image at medium quality. Token counts are the usage OpenAI returned; output image tokens were identical on repeat runs. Costs are those token counts multiplied by the list prices on OpenAI's pricing page as of the same date. GPT Image 2 numbers come from the same key, day and prompt. We did not test portrait, xhigh or max. Token counts measure billing, not quality, and OpenAI can change its prices or token accounting at any time.

Want to try Flare on your own prompts? Create a TulexAI account and pick it from the image model list on the Pro plan or above.

Frequently Asked Questions

What is GPT Image 2.5 Flare?

GPT Image 2.5 Flare is an OpenAI image model (API id gpt-image-2.5-flare, default snapshot gpt-image-2.5-flare-2026-09-08) that takes text and images as input and returns an image. OpenAI describes it as "Fast, high-quality everyday image generation" and offers it through the image generation and image edits endpoints.

How much does a GPT Image 2.5 Flare image cost?

OpenAI charges per token: $30 per million output image tokens. On our key on September 15, 2026, a high-quality 1024x1024 image billed 1,756 output tokens, about $0.053 including a short prompt, and medium billed 439 (about $0.013). On TulexAI it is 63 PT per image at high quality.

Is GPT Image 2.5 Flare cheaper than GPT Image 2?

Per token the prices are identical, but at medium and high GPT Image 2 billed 4x Flare's output tokens on our key, so a high-quality 1024x1024 image was about $0.211 vs $0.053 at list price. Low was identical at 196 tokens. On TulexAI both cost 63 PT per image: Flare runs at its high setting and GPT Image 2 at its medium setting, which bill the same 1,756 output tokens for a square image. See the full GPT Image 2 vs GPT Image 2.5 Flare comparison.

What is the difference between GPT Image 2.5 Flare and Sunburst?

OpenAI's image generation guide says to choose Sunburst for workflows where editing precision matters most, and Flare for fast, high-quality everyday image generation. We add no ranking beyond that guidance. Of the two, TulexAI offers Flare only.

Can I edit photos with GPT Image 2.5 Flare?

Yes. OpenAI documents the image edits endpoint for Flare, including reference images and masks. On TulexAI the edit path is attaching a photo with your prompt, and the edit runs on Flare at high quality; masks are not exposed. An edit with several reference images or a very long prompt counts as more than one image.

Does GPT Image 2.5 Flare support transparent backgrounds?

OpenAI documents transparent backgrounds for the 2.5 models in its image generation guide. TulexAI does not expose that option: every Flare request returns one opaque PNG.

Which TulexAI plans include GPT Image 2.5 Flare?

Pro ($23/mo) and every plan above it, at 63 PT per image. The $11 Basic plan does not include Flare or GPT Image 2; its image engine is GPT-5 Image Mini.