NewGPT Image 2.6 is live

GPT Image 2.6 online editor and generator

OpenAI's newest image model, released on 4 November 2026, in an editor you can use without an API key. Edit a photo with words, turn a photo or a sentence into a looping GIF, or make a PNG with a transparent background.

5 credits an image at 1K, GIFs from 4 credits. Credits never expire.

A still photo of white sneakers on a limestone block on the left, and on the right the same sneakers turning once round on the block, as a looping GIF made by GPT Image 2.6

GPT Image 2.6 on our three benchmark photos

Three fixed tests: the same three photos and the same words on every model page, so you can set one model beside another. GPT Image 2.6 Flare put the portrait on a palm beach with the face unchanged, removed both strangers, the scooter and the bin from the street photo, and moved the sneakers onto clean white. 13 to 17 seconds each.

OriginalGPT Image 2.6 Flare
Portrait: a new background
OriginalGPT Image 2.6 Flare
Product photo on white
OriginalGPT Image 2.6 Flare
Remove people and objects

What GPT Image 2.6 adds

Real runs from 5 November 2026, each with the words that made it. Use this prompt loads the photo, the words and the settings into the editor above.

GIF from a photo

A photo of sneakers that turns once round

We made the starting photo with GPT Image 2.6 too: new white sneakers on a limestone block, soft studio light. Then we asked for them to turn slowly through a full circle with the camera fixed. The editor's default GIF (9 frames, low quality) came back in 19 seconds: the shoes go once round with the same laces, the same sole and the same light, and the block and the backdrop stay put, so it loops cleanly. Use your own product photo the same way.

PromptThe sneakers turn slowly on the pedestal through a full circle, soft studio light, fixed camera, a smooth seamless loop. Keep the shoes exactly the same.

9 frames
19 seconds
6 credits

GIF from text

A looping mascot from one sentence

No photo this time: one sentence for a cartoon corgi that waves and blinks. All nine frames show the same dog, the same collar and the same outline, so the loop reads as one character moving. It took 19 seconds. Pick 6, 9 or 12 frames and Slow, Normal or Fast; the price shows next to the button before you start.

PromptA cute cartoon corgi mascot waving one paw and blinking, flat sticker illustration, plain white background, fixed camera, smooth seamless loop

9 frames
19 seconds
6 credits

A looping GIF of a cartoon corgi mascot waving one paw and blinking on a white background, made by GPT Image 2.6

Transparent PNG

A sticker with nothing behind it

Choose Background: Transparent and the file comes back as a PNG with real transparency. We asked for a die-cut avocado sticker with a thick white outline and checked the result on a checkerboard: outside the white edge, nothing is left, no glow or halo. 12 seconds.

PromptA die-cut sticker of a smiling cartoon avocado with a thick white outline

Transparent PNG
12 seconds
5 credits

A smiling cartoon avocado sticker with a white outline on a grey and white checkerboard that shows the background is transparent, made by GPT Image 2.6

Text in the picture

Both lines of the neon sign, spelled as written

The prompt asked for "OPEN LATE" in glowing cursive with a smaller line underneath. GPT Image 2.6 spelled both lines correctly, kept the capitals we wrote, and added nothing we did not ask for: no icons, no extra words. GPT Image 2.5 changed the case and added a coffee cup on the same prompt, so for exact wording this is the steadier model. 8 seconds at 1K.

PromptA neon sign on a dark brick wall that reads "OPEN LATE" in glowing cursive letters, with smaller white text underneath that reads "Coffee and cake until midnight". Photo, night, shallow depth of field.

Both lines correct
8 seconds
5 credits

A pink neon sign reading OPEN LATE in cursive with the line Coffee and cake until midnight underneath on a dark brick wall, made by GPT Image 2.6

Flare or Sunburst?

GPT Image 2.6 comes in two versions. Both cost the same here and both are in the model picker.

Same photo, same words

Sunburst kept the shoes at their size; Flare moved in close

On the product benchmark, Flare filled the frame with the shoes. Sunburst kept them at about the size they have in the original and left room around them. The portrait and the street photo came out alike in both. Flare was a little quicker (13 to 17 seconds against 16 to 19). Pick Flare for fast drafts and tight crops, and Sunburst when you want the object left at its original scale.

Flare 13 to 17 s
Sunburst 16 to 19 s
Same price

Worn sneakers on a cluttered floor on the left, then the same sneakers on white as made by GPT Image 2.6 Flare, large in the frame, and by Sunburst, smaller with more white space

Tips for better GPT Image 2.6 results

Say "keep the shoes exactly the same" or "keep her face unchanged" when the subject must stay. The model redraws the whole picture, so what you do not name can change.

Describe the loop for a GIF: the motion, a fixed camera, and "seamless loop". Nine frames at Normal speed is a good start; more frames cost more and take a little longer.

Use Fast for edits and drafts; Careful (same price) can help with crowded scenes or a lot of text.

Choose Transparent for stickers, logos and cut-outs; the file is a PNG. A transparent background is not combined with GIF.

How to use GPT Image 2.6

Three steps, no install.

Add a photo, or skip it
Upload up to 5 photos to edit them with GPT Image 2.6, or leave the tray empty to create an image from text.
Say what you want
Type the change in plain words. GPT Image 2.6 is already selected; the credit cost shows next to the button before you start.
Download or keep editing
The result appears under the tool and stays in My images for 30 days. Download it, or send it back in as the next photo.

GPT Image 2.6 compared with the other models

Same prompts, same portraits, measured on 2 November 2026 (GPT Image 2.6 on 5 November, on the benchmark photos).

Model	Credits at 1K	Seconds	Good at
Nano Banana 2.1	5	20 to 39	Everyday edits; keeps a likeness through a style change
Nano Banana 2	8	19 to 24	Speed; 16:9 and 21:9
Nano Banana Pro	10	30 to 36	Faithful faces; 4K
GPT Image 2	8	58 to 65	Lettering in a script; bold colour
GPT Image 2.5	8	36 to 64	Lettering; nine aspect ratios
GPT Image 2.6	5	8 to 20	Animated GIFs, transparent PNGs, exact lettering
Seedream 5.0 Pro	8 (12 at high)	29 to 93	Headshots closest to the original
Seedream 5.0 Lite	5	37 to 48	Cheap drafts, any quality
Grok Imagine	5	16 to 29	Fast, photographic scenes
Z-Image Turbo	1	13 to 19	Cheapest text-to-image drafts (no photos)

Seconds are wall-clock from pressing Generate to the finished image, over one to six runs per model; they vary with load. Credits are what this editor charges.

When to pick GPT Image 2.6

Pick it for animated GIFs, transparent PNGs and exact wording in a picture: no other model in this editor makes a GIF, and it kept our neon sign to the letter.

If you only need an everyday photo edit, Nano Banana 2.1 costs the same 5 credits at 1K; Seedream 5.0 Pro keeps a headshot closest to the original. GPT Image 2.5 stays available if you like its look.

AI Photo Editor pricing

A standard edit costs 5 credits; larger sizes and premium models cost 8 to 15. Buy a pack once and use it whenever you like. Bigger packs bring the price of a standard edit down from $0.17 to $0.12.

Starter

Try it out

$4.99one-time

30images

$0.17 per imageBase price

150 credits

Basic

Around 70 images

$9.99one-time

70images

$0.14 per imageSave 14%

350 credits

Plus

Around 150 images

$19.99one-time

150images

$0.13 per imageSave 20%

750 credits

ProRecommended

Around 240 images

$29.99one-time

240images

$0.12 per imageSave 25%

1,200 credits

Studio

For daily creators

$69.99one-time

600images

$0.12 per imageSave 30%

3,000 credits

Image counts and per-image prices here use 5 credits, what GPT Image 2.6 costs at 1K. The cost of the model and size you pick always shows before you start.

Secure one-time checkout through Stripe. Buying a pack only adds credits to your account. USD prices. Review the final amount and any applicable tax at checkout before paying. Credits do not automatically renew, and unused credits can be refunded within 7 days of purchase. The credit cost shows before each image; finished results need no second payment to view or download.

GPT Image 2.6 questions

What is GPT Image 2.6?

GPT Image 2.6 is OpenAI's image model released on 4 November 2026, after GPT Image 2.5. It comes in two versions, Flare for speed and Sunburst for detail. It creates images from text, edits photos from a written instruction, and can make animated GIFs and PNGs with a transparent background.

Is this the same as Microsoft's MAI-Image-2.6?

No. MAI-Image-2.6 is a different model from Microsoft. This page is about OpenAI's GPT Image 2.6.

How much does GPT Image 2.6 cost here?

An image costs 5 credits at 1K, 8 at 2K and 12 at 4K. A GIF costs 4 credits for up to 6 frames, 6 for up to 9 and 7 for up to 12. Credits come in one-time packs from $4.99 and never expire. A failed image costs nothing.

Is GPT Image 2.6 free?

No. A new account gets 5 free credits, which pays for one 1K image or one GIF of up to 6 frames.

How long does it take?

In our runs a Flare image took 8 to 20 seconds and a GIF 18 to 25 seconds. Sunburst took 15 to 19 seconds on edits.

What is the difference between Flare and Sunburst?

Flare is the faster one and Sunburst the one aimed at detail. Here they cost the same. On our product photo Flare zoomed in on the shoes while Sunburst kept them at their original size.

Can GPT Image 2.6 turn a photo into a GIF?

Yes. Add a photo, choose Format: Animated GIF, describe the motion, and pick 6, 9 or 12 frames and a speed. It redraws the picture, so expect a clean result in the style you describe rather than the same pixels.

Can it make a transparent background?

Yes. Choose Background: Transparent and the result is a PNG with real transparency. It cannot be combined with GIF.

Can it make SVG files?

Not here yet. We offer PNG, JPG-style stills and GIF.

Which aspect ratios does it support?

auto, 1:1, 3:2, 2:3, 4:3, 3:4, 16:9, 9:16 and 21:9.

How many photos can I upload?

Up to 5 per image.

Is this the official GPT Image 2.6 site?

AI Photo Editor is an independent site. It is not affiliated with, endorsed by or sponsored by the company that makes the model; model names belong to their owners.

Other models in the same editor

One account and one set of credits. Switch model from the model picker in the tool.

Try GPT Image 2.6 on your own photo

New accounts start with free credits. Failed images cost nothing.

Open GPT Image 2.6 See pricing