OpenAI · Released November 4, 2026

GPT Image 2.6 AI Image Generator

GPT Image 2.6 is OpenAI's newest image model: Flare for fast drafts, Sunburst for detail that holds across edits, and 50% faster generation than before.

This generator makes images with GPT Image 2.6 Flare.

Gallery

Real outputs from the OpenAI API, one image per prompt, no retouching.

Minimal jazz concert poster. Large headline "NIGHT SHIFT JAZZ", smaller lines "Saturday · 9 PM" and "Blue Lantern Club". Bold geometric shapes suggesting a saxophone and city lights, cream background with deep navy and mustard yellow, clean Swiss-style grid, crisp readable typography

Minimal jazz concert poster. Large headline "NIGHT SHIFT JAZZ", smaller lines "Saturday · 9 PM" and "Blue Lantern Club". Bold geometric shapes suggesting a saxophone and city lights, cream background with deep navy and mustard yellow, clean Swiss-style grid, crisp readable typography

Event posterGenerated with Flare
Product photo of a matte cream ceramic pour-over coffee kettle on a sand-colored stone block, soft morning window light, gentle shadows, a few coffee beans, minimal editorial styling, high detail

Product photo of a matte cream ceramic pour-over coffee kettle on a sand-colored stone block, soft morning window light, gentle shadows, a few coffee beans, minimal editorial styling, high detail

Product photoGenerated with Flare
Clean flat infographic titled "How a Banana Ripens". Five banana illustrations in a row from green to brown spotted, labeled "Day 1", "Day 3", "Day 5", "Day 7", "Day 9", each with one short caption below: "Firm", "Mild", "Sweet", "Very sweet", "Bake it". Soft cream background, consistent icon style, perfectly readable labels

Clean flat infographic titled "How a Banana Ripens". Five banana illustrations in a row from green to brown spotted, labeled "Day 1", "Day 3", "Day 5", "Day 7", "Day 9", each with one short caption below: "Firm", "Mild", "Sweet", "Very sweet", "Bake it". Soft cream background, consistent icon style, perfectly readable labels

InfographicGenerated with Sunburst
A café chalkboard menu photographed on a brick wall. Hand-lettered title "MORNING MENU" and five items with prices aligned on the right: "Oat Latte 4.5", "Banana Bread 3.8", "Avocado Toast 7.2", "Berry Yogurt 5.0", "Cold Brew 4.0". Small chalk doodles of coffee cups and leaves, natural light, realistic texture

A café chalkboard menu photographed on a brick wall. Hand-lettered title "MORNING MENU" and five items with prices aligned on the right: "Oat Latte 4.5", "Banana Bread 3.8", "Avocado Toast 7.2", "Berry Yogurt 5.0", "Cold Brew 4.0". Small chalk doodles of coffee cups and leaves, natural light, realistic texture

Chalkboard menuGenerated with Sunburst
Character turnaround sheet of a friendly banana mascot with round eyes, small smile and red sneakers: front view, side view, back view and a waving pose, all four poses identical in proportions and colors, flat vector illustration with thick dark outlines, white background, small label "BANANA BUDDY" at the top

Character turnaround sheet of a friendly banana mascot with round eyes, small smile and red sneakers: front view, side view, back view and a waving pose, all four poses identical in proportions and colors, flat vector illustration with thick dark outlines, white background, small label "BANANA BUDDY" at the top

Character sheetGenerated with Sunburst
Photoreal Scandinavian reading nook beside a tall window, rain streaks on the glass, a wool armchair, a warm brass floor lamp, stacked books and a sleeping cat, late afternoon blue light mixed with warm lamp light, interior design magazine photo

Photoreal Scandinavian reading nook beside a tall window, rain streaks on the glass, a wool armchair, a warm brass floor lamp, stacked books and a sleeping cat, late afternoon blue light mixed with warm lamp light, interior design magazine photo

Interior photoGenerated with Flare

Flare vs Sunburst

Flare or Sunburst: same prompt, two tiers

Both tiers share every feature and the same token price. Flare is tuned for speed; Sunburst for detail and consistency across edits.

Busy night market street in Hanoi after rain, red paper lanterns overhead, a glowing shop sign that reads "PHO 24H", scooters parked along the curb, dozens of people eating at low plastic stools, steam rising from bowls, reflections on wet pavement, cinematic photo
Generated with GPT Image 2.6 Flare

Flare — As little as about six seconds at quality low. Best for drafts, variations, posters and product shots.

Busy night market street in Hanoi after rain, red paper lanterns overhead, a glowing shop sign that reads "PHO 24H", scooters parked along the curb, dozens of people eating at low plastic stools, steam rising from bowls, reflections on wet pavement, cinematic photo
Generated with GPT Image 2.6 Sunburst

Sunburst — Higher quality and steadier edits at quality medium or high. Best for final assets and multi-round edits.

Busy night market street in Hanoi after rain, red paper lanterns overhead, a glowing shop sign that reads "PHO 24H", scooters parked along the curb, dozens of people eating at low plastic stools, steam rising from bowls, reflections on wet pavement, cinematic photo

What it does well

Lifestyle magazine cover. Masthead "SLOW" at the top in large serif letters, issue line "No. 26 · Winter", main cover line "The Art of Doing Less", three smaller cover lines on the left: "12 calm kitchens", "Walking as therapy", "Letters, not texts". A woman reading by a window in soft natural light, barcode in the bottom corner, elegant editorial layout

Lifestyle magazine cover. Masthead "SLOW" at the top in large serif letters, issue line "No. 26 · Winter", main cover line "The Art of Doing Less", three smaller cover lines on the left: "12 calm kitchens", "Walking as therapy", "Letters, not texts". A woman reading by a window in soft natural light, barcode in the bottom corner, elegant editorial layout

Generated with GPT Image 2.6 Sunburst

Layouts with real text

Headlines, cover lines and small print land where the prompt puts them, so posters and covers need fewer retries.

Detailed isometric cutaway illustration of a three-floor neighborhood bakery. Ground floor: shop counter with bread shelves and two customers. First floor: kitchen with an oven, a baker kneading dough and flour sacks. Top floor: small office with a desk and a cat on the windowsill. Small labels on each room: "SHOP", "KITCHEN", "OFFICE". Warm pastel colors, clean lines, many small consistent details

Detailed isometric cutaway illustration of a three-floor neighborhood bakery. Ground floor: shop counter with bread shelves and two customers. First floor: kitchen with an oven, a baker kneading dough and flour sacks. Top floor: small office with a desk and a cat on the windowsill. Small labels on each room: "SHOP", "KITCHEN", "OFFICE". Warm pastel colors, clean lines, many small consistent details

Generated with GPT Image 2.6 Sunburst

Busy scenes that stay organized

Many rooms, people and labels in one frame, each in its place. This is where the new complexity control helps most.

New in the API

What's new in GPT Image 2.6

Four new API capabilities. Image generation itself runs here, but these four controls are OpenAI API features and are not on nanobananana yet.

Low
Planning · Quick edits
Medium
Planning · Balanced
HighDefault
Planning · Dense layouts

More complexity means more planning before rendering.

Complexity

Not on nanobananana yet

Set how much the model plans before it renders: low for quick edits, high for dense text and layouts.

API details

Levels are low, medium and high; the default is high. It is separate from quality (low to max), can add latency, and has no separate fee.

Original
Markup layer
Result

“Recolor the circled vase to cobalt blue.”

Markup edits

Not on nanobananana yet

Circle the part to change on a transparent layer and describe the edit. The rest of the image stays as it was.

API details

Uses /v1/images/edits with supported GPT Image 2.5 and 2.6 models. The layer must match the first image's size, keep transparency, stay within 16,777,216 pixels and 50 MB, and cannot be combined with a mask. Same token rates as other edits.

Editable paths and anchor points

SVG output

Not on nanobananana yet

Get vector files with editable paths, shapes and colors for icons, logos and illustrations.

API details

Generations endpoint only, one image per request, no streaming or compression. The API renders a reference image first, then writes static SVG without scripts or embedded rasters.

9 frames × 200 msAbout 1.8 s per loop

GIF output

Not on nanobananana yet

Describe the subject and the motion, and get a looping animation from one prompt.

API details

Generations endpoint only. frame_count is 2 to 64 (default 9) and frame_duration_ms is 20 to 2,000 (default 200), so the defaults loop in about 1.8 seconds.

Overview

What is GPT Image 2.6?

GPT Image 2.6 is the model behind Images 2.6, released by OpenAI on November 4, 2026 in ChatGPT, Codex and the API. You still write a prompt or hand it images to edit; what changes is speed and control.

On nanobananana, GPT Image 2.6 Flare and Sunburst both run in the generator above and in the workspace, at 1K, 2K or 4K, with up to 16 reference images. The line under the prompt box names the tier that will make your image.

At a glance

Developer
OpenAI
Released
November 4, 2026
API models
Flare, Sunburst
Speed
50% faster than before
Generator here
Flare and Sunburst, 1K to 4K

Pricing

GPT Image 2.6 pricing

In the API, image output costs $30 per million tokens for both tiers, the same rate as GPT Image 2.5. On nanobananana, GPT Image 2.6 Flare and Sunburst cost the same: 3 / 4 / 7 credits per image at 1K / 2K / 4K, shown on the Generate button before you submit.

Credits per image on nanobananana
Model1K2K4K
GPT Image 2.6 Flare347
GPT Image 2.6 Sunburst347

Equal token rates do not mean equal cost per image: usage varies with tier, quality and size.

OpenAI API

$30 / 1M image output tokens

Comparison

GPT Image 2.6 vs GPT Image 2.5 vs Nano Banana Pro

Against GPT Image 2.5, the new version keeps the two tiers and the price but is faster and adds complexity, SVG and GIF. Nano Banana Pro is Google's model, a different family; run the same prompt in both to choose.

Show the comparison table
FeatureGPT Image 2.6GPT Image 2.5Nano Banana Pro
DeveloperOpenAIOpenAIGoogle
VersionsFlare, SunburstFlare, SunburstSingle model
On nanobanananaAvailableAvailableAvailable
Extra API featuresComplexity, markup, SVG, GIFMarkup—
Reference images hereUp to 16Up to 16Up to 8
Credits here (1K / 2K / 4K)3 / 4 / 73 / 4 / 78 / 8 / 12

How it works

How to use it on nanobananana

  1. Pick Flare or Sunburst
  2. Describe the image
  3. Add references, ratio and size
  4. Generate and keep editing

FAQ

GPT Image 2.6 FAQ

What is GPT Image 2.6?

OpenAI's image model released with Images 2.6 on November 4, 2026. It is in ChatGPT and Codex, and in the API as Flare and Sunburst.

Can I use it on nanobananana now?

Yes. GPT Image 2.6 Flare and Sunburst both run in the generator on this page and in the workspace, at 1K, 2K or 4K, with up to 16 reference images. Each image costs credits, shown on the Generate button. The API-only features (complexity, markup, SVG and GIF) are not offered here yet.

Flare or Sunburst: which should I use?

Flare for speed and drafts (as little as about six seconds at quality low). Sunburst for detail and consistency across edits at quality medium or high. Same features, same token price.

How much does GPT Image 2.6 cost?

$30 per million image output tokens in the API, the same as GPT Image 2.5. Here you pay credits per image, shown on the Generate button.

Can it make GIFs and SVG files?

Yes, through the API's generations endpoint (GIFs with 2 to 64 frames). Neither format is available on nanobananana yet.

What does the complexity setting do?

It sets how much the model plans before rendering: low for quick edits, medium or high for crowded scenes and dense text. The default is high and it costs nothing extra.

Who can use it in ChatGPT?

According to OpenAI, Images 2.6 is available globally in ChatGPT and Codex on every subscription tier, on desktop, mobile and web.

Ready to try it?

Write a prompt, pick Flare or Sunburst and generate.

Back to the generator
Share this page