OpenAI launches Images 2.6 with GIF animation, 50% faster generation

Key Points

1. OpenAI released Images 2.6 across ChatGPT, Codex and the API, adding animated GIF output and faster generation.
2. Users now create 4 billion images per week across OpenAI surfaces, more than half a billion per day.
3. Developers get two model tiers, image markup, SVG export and a new complexity control over editing.

The latest:

Images 2.6 is now live in ChatGPT, Codex and the API, generating images 50% faster than the previous version and adding animated GIFs for the first time. OpenAI said the release is built around editing consistency, keeping characters, products and brand details stable across successive rounds. Image creation across its surfaces now runs at 4 billion per week.

Details:

• The scale: OpenAI said people worldwide create 4 billion images per week across ChatGPT, Codex and the API, an average of more than half a billion images per day. The company framed the figure as the backdrop for a release aimed at cutting generation time rather than expanding access.
• The two models: In the API, developers choose between GPT-Image-2.6 Flare and GPT-Image-2.6 Sunburst. OpenAI said Flare can generate an image in as little as six seconds when quality is set to low, and improves layout for posters and product imagery. Sunburst targets higher-quality output at medium or high quality settings.
• The GIF feature: Images 2.6 can animate a still into a looping GIF, which OpenAI positioned for marketers animating brand mascots, social reactions and memes, and animated storyboards used to test character movement before video production. The company said a simple hand-drawn sketch is enough of a starting point.
• Editing consistency: OpenAI said the model holds characters, products and brand details steady as users edit, with Sunburst offering greater consistency across edits at medium or high quality. Consistency across successive rounds has been the main weak point of iterative image editing tools.
• New API controls: Image markup, previously available only in ChatGPT, moves to the API, letting teams point to the exact spot they want changed. SVG export is added so output can be used directly on websites, in apps and in print.
• Complexity parameter: A new complexity setting controls how much planning the model does before meeting a request, separate from the existing image quality setting that governs visual detail and fidelity. OpenAI recommends low complexity for faster edits and medium or high for scenes with many elements or constraints.
• Partner reaction: Coco Mao, CEO and co-founder of OpenArt, said GIFs are the native language of social media and that GPT-Image-2.6 lets creators produce them from a single prompt. Itay Schiff, creative director at Figma Weave, said the model is being added to Figma Weave as another option for creative production teams.
• Early deployment: Town, a startup building an AI work assistant, used Images 2.6 through the API to test animated characters in everyday workplace interactions, including an animated character named Monty that appears above the message box as a typing indicator.
• Pricing: OpenAI did not disclose pricing for either API model tier, nor did it say whether the GIF feature carries different billing from still image generation.

Between the lines:

The split between Flare and Sunburst, plus a complexity dial separate from quality, points at a cost-and-latency problem rather than a capability one: with half a billion images generated daily, OpenAI is giving developers a way to spend less compute on routine edits. The six-second low-quality figure is the clearest signal of that priority.

What’s next

Watch for OpenAI to publish API pricing for Flare and Sunburst, and for the Figma Weave integration to go live. Developer benchmarks on whether the 50% speed gain holds at medium and high quality settings will test the claim.