OpenAI's Images 2.6 makes GIFs and cuts generation times in half

Turning an idea into a quick animated reaction used to be an awkward multi-step chore: generate a static image, download it, import it into a third-party tool, and hope the animation loop holds together.

OpenAI is closing that gap. The company just rolled out Images 2.6 across ChatGPT, Codex, and its developer API, bringing native GIF creation directly into the conversation while cutting generation times in half.

The update is live globally across all subscription tiers on web, desktop, and mobile apps.

GIFs inside the chatbox

The marquee feature is native support for animated GIFs. You can now prompt ChatGPT to generate a moving loop in a single step—like asking for a dancing cat—and immediately tweak playback speeds or inspect frames inside the interface.

It works beyond text prompts, too. You can upload a photo of a product to turn it into an eye-catching motion loop, or feed the model a rough pencil sketch to create an expressive character animation for games or pitch decks.

Speed is the other headline. Images 2.6 generates visuals 50% faster than before. In API environments with lower quality settings, output can land in as little as six seconds.

Scale is scaling along with it. OpenAI says users now generate over 4 billion images per week across ChatGPT, Codex, and its API—averaging more than 500 million pictures a day. That is up from the 3 billion weekly images reported just two months earlier in September, according to The Verge.

Two API models and a complexity dial

For builders, OpenAI split the backend into two distinct API flavors: GPT-Image-2.6 Flare and GPT-Image-2.6 Sunburst.

Flare is the lightweight sprinter. When set to low quality, it generates images in about six seconds flat while polishing poster layouts and product imagery. Sunburst is the heavy lifter, built for medium-to-high fidelity and maintaining visual consistency across character iterations and repeated edits.

Developers also gain tools previously confined to the consumer interface. Image markup, which lets users pinpoint exact pixel regions for targeted revisions, is now an API endpoint. Designs can also be exported natively as scalable vector graphics (SVG) for web, app, and print workflows.

There is also a new complexity parameter. It lets developers dial up or down how much pre-generation planning the model commits to a prompt. Minor tweaks can run on low complexity for snappy iterations, while crowded scenes with intricate constraints can dial it up. Workplace AI startup Town is already tapping the API to inject animated characters into everyday work chats.

Why it matters

AI video gets all the cinematic hype, but full-blown video generation is still slow, expensive, and often excessive for everyday digital communication.

Between a frozen JPEG and a heavy video clip lies a massive, practical middle ground: the lightweight motion loop. By baking GIFs natively into the chat stream, OpenAI is capturing an everyday visual habit that users actually share across Slack, social media, and product mockups.

For creators and marketers, this bypasses timeline-based editing for quick branding assets, storyboards, and social memes. For engineers, splitting the offering into Flare and Sunburst—paired with a complexity slider—is a very pragmatic move. It lets developers choose between snappy sub-ten-second feedback loops and high-fidelity assets without paying a flat tax on latency.

Keep in mind that GIFs remain structurally limited in color depth and duration. This will not replace dedicated video models for cinematic narratives, but for quick-turnaround visual energy, it delivers exactly what was missing.