ReelyArt

AI image model · OpenAI

GPT Image 2.6 Sunburst, for detail and edits that hold.

GPT Image 2.6 Sunburst is the slower, more careful of the two image models OpenAI released on November 4, 2026, built for detailed work and editing. Hand it a photo and one instruction and it usually changes that and leaves the rest. It shares the Studio composer with Flare, GPT Image 2.5 and the rest of the catalog, at 1K, 2K and 4K on paid plans.

Open GPT Image 2.6 Sunburst in StudioAll models

4K

top resolution

3 credits

per image, from

8 images

reference cap

Paid plans

flagship tier

Edits

Change one thing, keep the rest.

Sunburst is built for editing. Each pair is one instruction on the frame before it, written as a change and a keep list. Drag the slider to compare.

The same coffee scene after a GPT Image 2.6 Sunburst edit printed a SLOW MORNING label onto the bagA white ceramic pour-over dripper on a glass carafe beside a kraft coffee bag with a blank label, before a GPT Image 2.6 Sunburst edit

Print a label onto a product shot

Three quoted strings set on a blank label. The dripper, the carafe, the beans and the bag's folds are the same frame.

Prompts

Edit

Print a label design onto the blank cream label on the coffee bag. Change only the label surface.
Text on the label, rendered cleanly and exactly once each, verbatim:
"SLOW MORNING" as the main title
"single origin · washed" beneath it
"No. 12" small at the bottom
Typography: tall condensed serif title in dark brown, small sans-serif lines beneath, one thin rule between them.
Keep the dripper, the carafe, the beans, the walnut board, the bag's shape and folds, the light, the camera angle and the framing exactly unchanged. No other text, no logos, no watermark.

Original image

Product photograph for a small coffee roaster's shop. A matte white ceramic pour-over dripper sits on a ribbed glass carafe on a dark walnut board, with a stand-up kraft paper coffee bag beside it. The bag has a large blank cream paper label on its front, completely empty. A few whole coffee beans scattered on the board. Soft window light from the left, warm neutral backdrop, shallow depth of field, natural colours. No text, no logos, no watermark.
The same reading corner turned into a snowy winter evening by a GPT Image 2.6 Sunburst edit, the floor lamp glowing and a knitted throw on the armchairA reading corner with a mustard velvet armchair by a sunny window on a summer afternoon, before a GPT Image 2.6 Sunburst edit

Summer afternoon to winter evening

Snow, dusk, a lamp switched on and a throw added. Our first try moved the camera and recoloured the chair; naming each object in the keep list held them.

Prompts

Edit

Edit this photo to a snowy winter evening. Change only the light, the view through the window and one added throw.
Through the window: falling snow, bare branches and a deep blue dusk sky.
Inside: the floor lamp switched on with a warm glow, evening light instead of sun, a chunky cream knitted throw draped over the armchair.
Preserve exactly: the camera position, focal length and framing; the mustard velvet armchair, its colour and shape; the round oak side table, the books and the cup; the fiddle-leaf fig and its pot; the picture on the wall; the window frame, walls and floorboards, all in the same place in the frame. Do not add or remove any furniture.

Original image

Interior photograph of a calm reading corner in a small apartment on a summer afternoon. A mustard velvet armchair beside a tall window, a round oak side table with a stack of books and a ceramic cup, a floor lamp switched off, a large potted fiddle-leaf fig, pale plaster walls, wide oak floorboards. Through the window, green trees in full leaf and bright sun casting soft window shadows across the floor. Wide 24mm architectural photography, eye-level, natural colours.
The same portrait after a GPT Image 2.6 Sunburst edit turned the bomber jacket burnt orange #C2410CA young man in an olive bomber jacket on a tram platform in late sun, before a GPT Image 2.6 Sunburst edit

One garment to an exact hex

The jacket moves to #C2410C and keeps its sheen and folds. His face, the tote and the platform do not move.

Prompts

Edit

Change the bomber jacket's colour to burnt orange #C2410C, keeping its nylon sheen, seams and folds. Keep the man's face, expression, hair, pose, the white tee, the tote bag, the tram platform, the background and the light exactly unchanged.

Original image

Street portrait of a young man in his twenties standing on a tram platform in the late afternoon, wearing an olive-green nylon bomber jacket over a white tee, a canvas tote on one shoulder, looking just past the camera with a relaxed half smile. Short curly dark hair. Behind him a blurred tram, a timetable board with no legible text and warm low sun. 50mm, eye-level, shallow depth of field, natural skin texture, subtle film grain.

Step by step

One product, three edits deep.

Each step edits the previous result: a new flavour, then a new scene. The tin, the type and the wordmark carry through.

The same tea tin turned sage green and relabelled Jasmine Green by a GPT Image 2.6 Sunburst editA matte navy tea tin labelled KESTREL Earl Grey on linen beside a cup of tea, made with GPT Image 2.6 Sunburst

Edit 1, made on step 0

Make this a new flavour of the same tea. Change the tin's colour from navy to sage green #6B8F71. Change the text "Earl Grey" to "Jasmine Green" in the same italic serif, size and position. Replace the cornflower petals with a few small white jasmine flowers. Keep "KESTREL", "loose leaf · 100 g", the tin's shape, the label, the cup, the linen, the light and the camera exactly unchanged.
All prompts

Step 0, the first picture

Product photograph of a round tea tin on a pale linen surface. The tin is matte navy blue with a cream paper label wrapped around it. Label text, rendered cleanly and exactly once each:
"KESTREL" as a small-caps wordmark at the top
"Earl Grey" in a large italic serif in the middle
"loose leaf · 100 g" small at the bottom
A small porcelain cup of tea and a few dried cornflower petals beside the tin. Soft diffuse daylight from the upper left, gentle contact shadow, warm neutral backdrop. 70mm, eye-level, centred. No other text, no watermark.

Step 1, a new flavour

Make this a new flavour of the same tea. Change the tin's colour from navy to sage green #6B8F71. Change the text "Earl Grey" to "Jasmine Green" in the same italic serif, size and position. Replace the cornflower petals with a few small white jasmine flowers. Keep "KESTREL", "loose leaf · 100 g", the tin's shape, the label, the cup, the linen, the light and the camera exactly unchanged.

Step 2, a new scene

Place the same sage green KESTREL Jasmine Green tin from the image on a sunlit wooden windowsill in the morning, beside a glass teapot with steeping leaves and a small potted herb. Soft golden side light, the garden blurred through the window. Keep the tin's shape, colour, label design and every word on it exactly as they are, at a believable scale for the scene. No other text, no watermark.

Reference images

A person, a garment and a place.

Three reference images, each named by its role in the prompt, composed into one street photo. The inputs were made on Flare.

  • A studio portrait of a young woman with wavy auburn hair and freckles, used as reference image 1The person
  • A flat lay of a mustard cable-knit cardigan with wooden toggles, used as reference image 2The garment
  • A steep old-town street with blue and white tiled facades in morning sun, used as reference image 3The place
The woman from the portrait walking up the tiled street in the mustard cardigan, composed by GPT Image 2.6 Sunburst from three references

Three references, one photo

Her face and freckles from the portrait, the cardigan's cable knit and toggles from the flat lay, the tiles and light from the street.

Prompt
Image 1: the person. Image 2: the cardigan. Image 3: the location.
Photograph the woman from Image 1 walking up the tiled street from Image 3, wearing the mustard cardigan from Image 2 over a white t-shirt with light blue jeans, mid-stride, glancing back toward the camera with a small smile.
Keep her face, hair and freckles exactly as in Image 1, the cardigan's colour, cable knit and wooden toggles exactly as in Image 2, and the tiled facades, cobblestones and morning light of Image 3.
Candid street-style photograph, 50mm, eye-level, shallow depth of field, natural skin and fabric texture. No text, no logos, no watermark.

Detail

Where the extra seconds go.

Texture, small type and dense scenes, each from a single prompt at 1K.

A macro of an open mechanical watch movement with brass gears, ruby jewels and steel tweezers, made with GPT Image 2.6 Sunburst
Brass, jewels and steelA macro of a watch movement: perlage on the plates, a blued screw, tweezers holding one gear.
Prompt
Macro photograph of an open mechanical watch movement on a watchmaker's bench: brass gear train, red synthetic ruby jewels, a blued steel screw, a balance wheel mid-swing, and the tip of a pair of steel tweezers holding a tiny gear above it. Fine perlage and brushed finishing on the plates, a loupe and a soft blue bench mat blurred behind. Focused, cool overhead light with one warm accent, very shallow depth of field. No brand names, no text, no watermark.
A vintage botanical plate titled Studies of Ferns with six numbered watercolour fern fronds, made with GPT Image 2.6 Sunburst
A plate of fine printSix ink-and-watercolour ferns, numbered, under a title and two small lines of type.
Prompt
A vintage-style botanical study plate printed on aged cream paper. Six fern fronds drawn in fine ink line with soft watercolour washes in muted greens, arranged in a neat two-column grid, each with a small numeral beneath it: "1" "2" "3" "4" "5" "6".
Text, rendered cleanly and exactly once each:
"PLATE XII" small and centred at the top
"Studies of Ferns" as the title beneath it, in an elegant serif
"Drawn from life · Spring" small at the bottom
Fine paper grain, faint foxing at the corners, a thin printed border. Straight-on flat scan. No other text, no watermark.
An isometric cutaway of a two-storey corner bakery signed CRUMB & CO with bakers at work and a flat upstairs, made with GPT Image 2.6 Sunburst
A busy little buildingAn isometric bakery with bakers, a customer, a flat upstairs and one shop sign.
Prompt
Isometric cutaway diorama of a two-storey corner bakery: on the ground floor a counter with glass cases of bread and pastries, a wood-fired oven glowing, two bakers shaping dough at a floured table and one customer at the counter; upstairs a small flat with a sofa, a bookshelf and a cat on the windowsill; flour sacks stacked by the back door and a bicycle outside.
Text, rendered exactly once, on the small sign above the shop door: "CRUMB & CO"
Polished miniature 3D render, true isometric projection with parallel edges, warm morning light, soft ambient occlusion, cream void background with a soft shadow under the block. No other text, no logos, no watermark.
A cut sourdough loaf with an open crumb and a blistered crust on floured linen, made with GPT Image 2.6 Sunburst
Crumb and crustGlossy open crumb, a blistered crust and rice flour along the score, in low raking light.
Prompt
Close-up food photograph of a freshly cut sourdough loaf on a floured linen cloth: the open crumb with irregular glossy holes, a deeply caramelised blistered crust with a dusting of rice flour along the score, one slice leaning against the loaf, a serrated knife and a small dish of salted butter at the edge of frame. Low raking window light from the right, dark rustic wooden table, 85mm, shallow depth of field. No text, no watermark.

Sunburst vs Flare

One prompt, both 2.6 models.

The same prompt and size on each. Sunburst takes longer over the detail; Flare comes back sooner.

A crowded stall with handwritten tags

Both set the sign and all three price tags as written. Flare filled the table edge to edge; Sunburst framed a looser stall with more of the market behind it.

A flea-market stall of vintage film cameras with handwritten price tags, made with GPT Image 2.6 Sunburst from the same prompt
GPT Image 2.6 Sunburst
A flea-market stall of vintage film cameras with handwritten price tags, made with GPT Image 2.6 Flare from the same prompt
GPT Image 2.6 Flare
Prompt
A flea-market stall of vintage film cameras at a Sunday market: rows of old rangefinders, folding cameras and twin-lens reflexes on a green felt table, small handwritten paper price tags tied with string reading "$40", "$65" and "$120", a cardboard sign reading "FILM TESTED", a vendor's hands polishing a lens in the corner. Soft overcast light, 50mm, eye-level, natural colours. No brand names on the cameras, no watermark.

Glaze, crackle and rain

Both kept the speckled glaze, the crackle and the rain. Flare's fox sits alert and looks out; Sunburst's smiles with its eyes closed, the glaze pooling heavier at the base.

A glazed ceramic fox figurine on a rainy windowsill, made with GPT Image 2.6 Sunburst from the same prompt
GPT Image 2.6 Sunburst
A glazed ceramic fox figurine on a rainy windowsill, made with GPT Image 2.6 Flare from the same prompt
GPT Image 2.6 Flare
Prompt
A glazed ceramic fox figurine sitting on a rainy windowsill: speckled orange and cream glaze with drips pooled at the base, a hairline crackle in the glaze, raindrops and streaks on the window glass behind, a blurred grey street beyond. Soft cool daylight with one warm reflection on the glaze. 85mm, macro-close, shallow depth of field. No text, no watermark.

GPT Image 2.6 Sunburst vs GPT Image 2.6 Flare

Two models, different jobs.

Sunburst is the one you finish on. Both take the same sizes and cost the same credits, so nothing is lost by drafting on Flare and bringing the winner here.

 

GPT Image 2.6 Sunburst

GPT Image 2.6 Flare

Built for

Detailed work and editing

Fast iteration

Speed

Slower, more deliberate

The faster of the two

Reach for it when

A face, a product or small text has to survive the edit

You want several directions or a set that matches

Cost here

3 / 4 / 6 credits

3 / 4 / 6 credits

Not sure? Start here and move over when a detail has to survive. See what GPT Image 2.6 Flare is for.

Why this model

What GPT Image 2.6 Sunburst is for.

Edits that hold

Name the one thing that changes and say what stays: the face, the folds, the light. The label, season and colour edits on this page were one instruction each.

Detail worth the wait

It takes longer than Flare and spends the time on texture and small type: watch gears, a sourdough crumb, crackled glaze, a plate of fine print.

Several references, one scene

Up to 8 images, each given a role: a face, a garment, a place. The street photo below was built from three references and one prompt.

Same sizes and price as Flare

3 credits at 1K, 4 at 2K, 6 at 4K, the same as Flare, so the slower render never costs more.

Prompt guide

How to prompt GPT Image 2.6 Sunburst.

Sunburst follows a brief written in parts. For an edit, the part that matters most is the list of what must not move.

The shape

Change: the one thing that changes
Details: exact colour, quoted words, placement
Keep: everything that must not move
Leave out: extra text, logos, new objects

A worked example

Make this a new flavour of the same tea.
Change the tin from navy to sage green
#6B8F71. Change "Earl Grey" to "Jasmine
Green" in the same serif, size and place.
Keep "KESTREL", the tin's shape, the cup,
the linen, the light and the camera
exactly unchanged.

Write the keep list

Name what must not move: face, pose, light direction, camera angle, framing. Then name the one change. Restate the list on each follow-up, or an edit drifts into a redraw.

Quote every word you need

Put each string in quotation marks and say where it goes and that it appears once. A described caption comes back paraphrased; a quoted one comes back as written.

Give every reference a role

With several images, say which is the person, which the garment, which the place. "Image 1: the person" is enough.

Colours by hex

An exact hex code holds a brand colour better than "dark orange". Ask for the material to stay too: "keeping its sheen and folds".

Use cases

Where GPT Image 2.6 Sunburst fits.

Colourways and flavours

One product shot, every colour and label, by hex code and quoted text.

Labels and packaging

Print a design onto a blank label without reshooting the product.

Seasons and time of day

The same room or street in winter, at night, in the rain.

Campaign composites

A person, a garment and a location from separate references in one frame.

Final assets

The detailed frame that ships, after the idea was found on a faster model.

Pricing

What one image costs.

1KPaid plans3credits
2KPaid plans4credits
4KPaid plans6credits

Credits per image; four variations cost four times the number shown. Aspect ratios: 1:1 · 4:5 · 3:4 · 9:16 · 16:9 · 2:3 · 5:4 · 4:3 · 3:2 · 2:1 · 1:2 · 21:9 · 9:21 · 3:1 · 1:3. Failed generations are refunded automatically. Plans and credit packs are on the pricing page.

FAQ

Before you generate.

Last verified 5 Nov 2026

What is GPT Image 2.6 Sunburst?

One of two image models OpenAI released on November 4, 2026. Sunburst is the one built for detailed work and editing; its sibling, GPT Image 2.6 Flare, is built for fast iteration.

Is GPT Image 2.6 Sunburst free?

It is on paid plans, at every size. On the free plan, GPT Image 2.5 Sunburst runs at 1K and is the closest model to it. No OpenAI account or API key is involved either way.

Sunburst or Flare?

Same price, same sizes, same reference cap. Sunburst is slower and more careful; Flare is the quick one. Draft on Flare, then re-run or edit on Sunburst when the detail matters.

How is it different from GPT Image 2.5 Sunburst?

It is OpenAI's newer generation of the same careful model, and it costs more to run per image, so it costs more credits here: 3 / 4 / 6 credits against 2 / 3 / 4 credits for 2.5. Both stay in Studio; run one prompt on each and keep the one you prefer.

How do I edit a photo with it?

Upload the photo in Studio, pick GPT Image 2.6 Sunburst and write the change first, then what must stay: "change the jacket to burnt orange #C2410C, keep his face, pose and the background unchanged". Restate the keep list on every follow-up edit.

What does one image cost?

3 credits at 1K, 4 at 2K, 6 at 4K, an edit included. Failed generations are refunded automatically.

Generate with GPT Image 2.6 Sunburst.

Open Studio
All modelsGPT Image 2.6 FlareKrea 3Nano Banana 2.1Ideogram 4.5FLUX 3GPT Image 2.5 FlareGPT Image 2.5 SunburstNano Banana 2Nano Banana ProGPT Image 2GPT Image 1.5Seedream 5 ProSeedream 5 LiteSeedream 4.5Grok Image 2AI StudioTemplatesGuidePricing