OpenAI's Images 2.6 brings animated GIFs to ChatGPT and 50% faster images

OpenAI released Images 2.6 on November 4, 2026, and says it generates images 50% faster than before. The update also teaches ChatGPT to make animated GIFs from a prompt, a photo or a rough sketch. It is live in ChatGPT, Codex and the API, where developers get two new models called Flare and Sunburst.

What is Images 2.6?

Images 2.6 is OpenAI's new image generation update, and it ships to all three of its surfaces at once. The announcement frames it around two goals: faster results and more control over editing rounds.

In practice, that means keeping characters, products and brand details consistent while you keep refining a picture. OpenAI's pitch names marketers and creators in particular.

In the API, the release is split into two models. GPT-Image-2.6 Flare is the fast one. GPT-Image-2.6 Sunburst is the one for detail. Both follow the 2.5 versions of the same names, which OpenAI's image guide still uses in its code samples.

The scale behind the launch is large. OpenAI says people now create 4 billion images per week across ChatGPT, Codex and the API, or more than half a billion a day. The Verge notes that the company reported 3 billion per week in September.

How do GIFs work in ChatGPT?

You ask for one in the chat, and ChatGPT returns a looping animation. According to The Verge, you can start from a prompt or an uploaded photo. You can then adjust the animation speed, edit individual frames and share the result.

That is new for ChatGPT. Before this update, it couldn't natively generate a GIF in a single step, The Verge reports. In an example OpenAI shared, a user types "can you make a gif of a cat dancing?" and gets controls for playback speed and a frame-by-frame view.

OpenAI also says a drawing is enough to start. Even a simple sketch can become an animated character for videos, games and more. The company lists the uses it has in mind:

• animating a brand mascot for a campaign;
• turning a product shot into a short loop;
• making reactions and memes for social media;
• building animated storyboards to test character movement before video production.

OpenArt, an image platform whose examples appear in the announcement, is quoted there too. "GIFs are the native language of social media," said its CEO, Coco Mao. So far, people who wanted this kind of loop often reached for separate animation tools. Now a first draft can come straight from the chat window.

How fast is Images 2.6, and how do Flare and Sunburst differ?

OpenAI says Images 2.6 generates images 50% faster than before. It doesn't say which earlier model or which settings it measured against, however.

The two API models trade speed for detail:

• Flare can return an image in as little as six seconds when quality is set to low. OpenAI adds that it improves layout for posters, product images and other professional visuals.
• Sunburst aims for higher quality and greater image consistency across edits when quality is set to medium or high.

So the split is simple. Flare suits quick drafts and interactive apps, because a user waits for each result. Sunburst suits work where the same face, product or logo has to survive several rounds of edits. If character consistency is your main problem, Sunburst is the model OpenAI pitches for consistency across edits.

No independent speed tests were public when we wrote this. The Verge repeats the speed claim with attribution to OpenAI, and so do we.

What's new for developers in the API?

Developers get four new controls on top of the two models. OpenAI's image generation guide says both Flare and Sunburst support all of them through the Image API:

• Image markup. Already in ChatGPT, it now comes to the API, so teams can mark exactly where a change should go.
• SVG export. Designs come out as vector files for websites, apps and print.
• Animated GIF output. The same GIF feature as in ChatGPT, available to apps.
• Complexity. A new setting that controls how much the model plans before it renders.

How does the complexity setting work?

Complexity takes three values: low, medium and high. If you leave it out, both models use high. Lower settings suit focused edits, such as swapping a background or a sign. Higher settings help with dense text, complex layouts and scenes with many objects.

There is a catch, though. The guide warns that higher complexity can increase latency and introduce unwanted changes. For most edits, OpenAI recommends low complexity paired with the image quality you need. Complexity has no separate fee.

Quality and complexity are separate dials. Quality controls visual detail and fidelity, while complexity guides how much goes into the scene. Figma Weave, a design tool, says it plans to add the model to its workflows.

How much do Flare and Sunburst cost?

Both models cost the same per token, and the rates match the 2.5 generation. OpenAI's Flare model page lists $30 per million image output tokens and $8 per million image input tokens. The Sunburst page shows the same numbers.

OpenAI doesn't give a price per image on these pages. The bill depends on how many tokens each image uses. So Flare's advantage is time, not a cheaper rate.

One detail is easy to miss. The 2.6 model pages list the Batch API as not supported. By contrast, the 2.5 pages quote a Batch rate of $15 per million image output tokens, half the standard price. Teams that render large jobs overnight may want to stay on 2.5 until that changes.

Who can use Images 2.6 now?

Every ChatGPT and Codex subscription tier should have it. OpenAI says Images 2.6 is "available globally today in ChatGPT and Codex across all subscription tiers on desktop, mobile, and web." The Verge describes it as rolling out to all ChatGPT and Codex users.

The API models are live too, under the IDs gpt-image-2.6-flare and gpt-image-2.6-sunburst. They work with the image generation and image edit endpoints.

For a ChatGPT subscriber, nothing needs switching on. For a developer, the move is a model ID change, plus a decision on complexity. If you're comparing it with other text-to-image tools, the cards below show each one's current plans.

C

ChatGPT

70

General assistant with tools, memory, files and voice

🇺🇸United States·Added 2mo ago

Freemium·from $8/mo·API

89trust score out of 100

2

O

OpenAI Codex

91

OpenAI's coding agent for the terminal, IDE, desktop and cloud sandboxes, included with ChatGPT plans or billed by API tokens.

🇺🇸United States·Added 27d ago

Freemium·from $8/mo·API

94trust score out of 100

1

M

Midjourney

49

Subscription image generator with a distinctive illustrative house style

🇺🇸United States·Added 2mo ago

Paid

86trust score out of 100

4

A

Adobe Firefly

68

Adobe's generative image, video and audio studio with its own commercially safe models plus partner models.

🇺🇸United States·Added 2mo ago

Freemium·API

90trust score out of 100

3

R

Recraft

71

Raster and editable vector generation with reusable brand styles, a design canvas, Figma plugin and API

🇬🇧United Kingdom·Added 2mo ago

Freemium·from $10/mo·API

91trust score out of 100

I

Ideogram

60

Text-to-image model that gets words inside the picture right

🇨🇦Canada·Added 2mo ago

Freemium·from $8/mo·API

85trust score out of 100

What we don't know yet

The launch answers the what, but several gaps remain:

• The baseline. "50% faster" has no named reference model or setting.
• GIF limits. OpenAI gives no maximum length, frame count, resolution or per-plan caps for GIFs.
• Free accounts. The wording says "all subscription tiers" and doesn't spell out limits for each plan.
• Per-image cost. The API is priced per token, with no per-image figure for either model.
• Safety. OpenAI points to a system card for its safety evaluations. We have not seen outside testing of the GIF feature yet.

We read OpenAI's announcement, both model pages and the image guide side by side, then checked them against The Verge's report. Our view: for most ChatGPT users, GIFs are the headline, and they are already in the app. For developers, the real changes are SVG export and the complexity setting. The prices stay flat, but the missing Batch support matters for bulk work.

Next, watch for independent speed tests of Images 2.6 and for OpenAI's help pages on GIF limits. For Business plans, OpenAI says usage will share one limit across ChatGPT and Codex from December. It doesn't yet say how image and GIF generation will count against that limit.