Product Launch

GPT Image 2.6 Released: New Features and API Updates

Discover GPT Image 2.6’s release: faster generation, more consistent edits, complexity controls, markup, GIF animation, SVG output, and API availability.

EvoLink TeamProduct Team

November 5, 20266 min read

On this page

GPT Image 2.6 is OpenAI's image generation and editing model family, released on November 4, 2026, with Flare and Sunburst. It can create images from text and revise existing images. This release adds animated GIF and SVG output, introduces complexity to control planning separately from quality, and adds markup-guided editing: circle an object and describe how to change it. OpenAI also reports faster image generation and better consistency across edits. Official release

Creators and developers can use these updates for product and campaign images, targeted edits, short animations, and vector graphics.

GPT Image 2.6 release: what is new and improved?

Release update	What it does	A concrete use
Faster image generation	OpenAI reports images generating 50% faster than before	Produce and revise visual ideas with a shorter wait
Better editing consistency	OpenAI reports improved preservation of character, product, and brand details across edits	Revise a product scene while keeping its identity recognizable
Complexity control	Adjust planning effort separately from rendering quality	Give a layout with several objects and text more planning
Markup-guided editing	Use a visual annotation and a written instruction to identify a change	Circle a vase and request a new color
Animated GIF output	Generate a short animation from text or still references	Animate a mascot or turn a product image into a loop
SVG output	Produce vector graphics through OpenAI's generation interface; not yet available on EvoLink	Create icons or illustrations with editable shapes and colors

Markup is part of the release and is also available on supported GPT Image 2.5 models. Release notes

Faster image generation and more consistent edits

OpenAI says Images 2.6 generates images 50% faster than before. In the API, Flare can generate an image in as little as six seconds when quality is set to low. OpenAI announcement

Successive edits now focus on preserving character, product, and brand details. OpenAI reports better detail and consistency with Sunburst at medium or high quality, along with improved poster and product-image layouts with Flare.

These improvements apply to product visuals, campaign assets, characters, and layouts containing text.

Complexity and markup: new controls for creating and editing images

Complexity controls planning separately from quality

The new complexity setting controls how much planning the model puts into composition, text, and relationships between objects before generating an image. quality controls rendering quality, so the two can be set independently. Both 2.6 models offer low, medium, and high complexity, with high as the default. Image generation guide

A focused change, such as replacing a background, may need less planning. A poster with several objects, an exact headline, and relationships between elements may benefit from more.

low complexity: red jerseys, a crowned trophy, and a foreground banner

medium complexity: red jerseys, a laurel trophy, and a foreground banner

high complexity: blue jerseys and CHAMPIONS lettering in the stands

OpenAI-published Sunburst examples using the same prompt and medium quality, ordered from low to high complexity. Notice how the trophy, uniforms, and lettering placement change. Source.

Markup lets you point to the object you want to change

Markup lets you circle an object, such as “the vase on the table,” on a separate transparent PNG layer and describe the change in the prompt. The model uses the visual annotation and text together to identify what to change. Markup editing

OpenAI's example below circles a striped vase and asks Flare to make it blue. The annotation tells the model where to look; the written instruction tells it what to do.

Markup input: a red circle identifies the striped vase

Flare output: a blue vase with the red annotation removed

OpenAI's markup example: annotated input and Flare output. The vase becomes blue and the red annotation is removed. Source.

GIF and SVG output: animation and vector graphics

GIF turns a prompt or still reference into a short animation

With GPT Image 2.6, you can now create short animations alongside still images: a mascot waving, a character reacting, or a product moving in a loop.

OpenAI's image generation and image editing APIs support GIF output. The editing API can use still reference images to guide the animation. Frame count and per-frame duration control the animation's sequence and timing. GIF documentation

Sunburst GIF: a monkey lifts a lantern and returns to its starting pose

OpenAI-published Sunburst example: a monkey lifts a lantern and returns to its starting pose. Nine frames at 200 ms each form a 1.8-second loop. Source.

For motion prompts, frame timing and file checks, follow the GPT Image 2.6 GIF tutorial.

SVG provides graphics that can be scaled and edited

SVG is a vector format: shapes, paths, and colors can be adjusted in a compatible design tool, and the graphic can be scaled for different placements. This makes it relevant to icons, simple illustrations, and other assets that need further design work.

OpenAI's image generation API supports SVG output; the editing API does not. SVG documentation

On EvoLink, both Flare and Sunburst support GIF output. SVG output is not available yet.

GPT Image 2.6 API: released models and EvoLink support

OpenAI released two API models: Flare, with the ID gpt-image-2.6-flare, and Sunburst, with the ID gpt-image-2.6-sunburst. Both support image generation and editing. Flare model card, Sunburst model card

On EvoLink, Flare and Sunburst support image generation, reference-image editing, complexity, markup, and GIF output. Try the models, view current pricing, and find API examples on their model pages. Both routes are available through EvoLink's unified API gateway.

OpenAI lists the same standard token rates for the 2.6 and 2.5 models. See EvoLink's model pages for current pricing. Official pricing

Already using 2.5? Compare quality tiers, output costs and migration choices for 2.6 vs 2.5. For selecting a 2.6 variant, see Flare vs Sunburst.

GPT Image 2.6 release FAQ

What is GPT Image 2.6?

It is OpenAI's image generation and editing model family, released on November 4, 2026, with Flare and Sunburst. This release improves generation and editing and adds complexity control, markup-guided edits, GIF animation, and SVG output.

Can GPT Image 2.6 animate an existing still image?

Yes. You can use a still reference image to guide GIF generation through OpenAI's image editing API, describing the movement and loop in the prompt.

Is complexity the same as quality?

No. Complexity controls planning before rendering; quality controls rendering quality. They are separate settings, so a more complex request does not automatically mean a higher-quality rendering tier.

Is markup exclusive to GPT Image 2.6?

No. Markup is also available on supported GPT Image 2.5 models.

Are GIF and SVG both available through EvoLink?

Both Flare and Sunburst on EvoLink support GIF output. SVG output is not available yet.

All Posts

#GPT Image 2.6 #OpenAI #Flare #Sunburst #Image Editing