ChatGPT Images 2.6 adds GIFs, and Meta may label the ads you make

On Wednesday, November 4, OpenAI introduced Images 2.6 in ChatGPT, Codex and its API. The headline number: images now come out 50% faster than before. OpenAI also says people create 4 billion images a week across those three products, which works out to more than half a billion a day. That's a lot of cats in sunglasses.

The update is out now in ChatGPT and Codex on every subscription tier, on desktop, mobile and web. In the API, developers choose between two models: GPT-Image-2.6 Flare, built for speed, and GPT-Image-2.6 Sunburst, built for detail. OpenAI says Flare can return an image in as little as six seconds when quality is set to low, and that it lays out posters, product images and other professional visuals better than before.

Sunburst is the careful sibling. At medium or high quality, OpenAI says it keeps characters, products and brand details more consistent from one edit to the next. Anyone who's watched a logo slowly melt by the third revision knows why that line matters.

The fun part is motion. Images 2.6 can turn a still image into a GIF, and OpenAI's own examples are aimed squarely at marketers: animate a brand mascot for a campaign, make reactions for social media, or map out a character's movement in an animated storyboard before you pay for video production. OpenAI says even a simple sketch can become an animated character, which is good news for everyone whose mascot currently lives on a napkin.

"GIFs are the native language of social media," said Coco Mao, the CEO and cofounder of OpenArt, in OpenAI's announcement. Your feed would agree.

For developers, the API gets three practical upgrades. Image markup, already in ChatGPT, lets a team point to the exact spot that needs a change instead of describing it in a paragraph. SVG export turns designs into files that stay sharp on websites, apps and print. And a new complexity setting controls how much planning the model does before it draws. OpenAI recommends low complexity for quick edits and medium or high for scenes with many elements.

The price didn't move. On OpenAI's API pricing page, both GPT-Image-2.6 models list the same per-token rates as the 2.5 models: $8 per million image input tokens and $30 per million image output tokens. Faster for the same money is the rare software update with no catch on the invoice. The catch is in the file itself.

Most people will skip this part of the release. OpenAI's system card for Images 2.6 says its images carry C2PA metadata, an industry standard for disclosing where a piece of content came from, plus an invisible SynthID watermark from Google DeepMind. That applies in ChatGPT, Codex and the API. The same card says GIF creation got its own additional safeguards.

Meta uses signals like these. Its Transparency Center says it adds an "AI info" label to posts when it detects industry standard AI image indicators, or when people disclose that they are uploading AI-generated content. For ads, Meta updated its approach on June 1, 2026: Meta is rolling out an "About this ad" menu that will hold AI info labels, and Meta says it will begin automatically detecting ads created or edited with third-party AI tools and label them there. Meta adds that the experience may vary by region because of legal requirements.

So a GIF made in ChatGPT and run as a Meta ad can arrive with its paperwork attached, and the platform can show that paperwork to the person scrolling past it.

Speed and consistency make AI images useful for the work that eats creative hours: social posts, mascot moments, storyboards and quick product layouts. Your brand rules still matter more than the model. Images in as little as six seconds means you can make 40 versions before lunch, and 40 versions of an off-brand image is still 40 off-brand images.

The labels change what AI imagery is good for. An animated mascot or an illustrated storyboard with an "AI info" tag is fine; nobody thinks your cartoon owl is a documentary. A realistic photo of a customer, a result, or a product doing something it doesn't do is a different story, labeled or not. Use AI for imagination, and use real photos for proof.

For a small team, the easiest win is motion in places that already get attention: one looping shot of a real product for your next social post, one mascot reaction your audience will actually reuse, or one animated storyboard to test a video idea before anyone books a camera crew.

When anyone can make a polished image in seconds, polish won't set you apart. A clear brand and real proof will, along with knowing which job belongs to which tool. Images 2.6 makes the polish cheaper and faster. The rest is still on you.

Want a second set of eyes on your creative before it ships? Talk to our founders below. See you in the next edition.