GIFs and 50% faster generation land in ChatGPT Images 2.6

OpenAI says Images 2.6 generates images 50% faster than before. In the API, its November 4 release splits the work between two new models, so the headline speed and the steadier edits come from different choices: GPT-Image-2.6 Flare gets down to as little as six seconds per image only with quality set to low, and GPT-Image-2.6 Sunburst holds details more consistently across edits at medium or high quality.

Beyond the API, OpenAI’s Images 2.6 announcement adds GIF creation and reusable reference images. The update is live globally in ChatGPT and Codex on every subscription tier, on web, desktop and mobile.

Picking the model and the settings

Flare is the speed option, and OpenAI says it also lays out posters, product imagery and other professional visuals better. Sunburst is for work that needs more detail and precision, with higher-quality output and more consistency from one edit to the next when quality is medium or high. The pair keeps the split from Images 2.5, where Flare was pitched as the default for most applications and Sunburst as the model for premium workflows that need tighter control across edits.

The API also gets a new complexity parameter, which sets how much planning the model does before it meets a request. Quality and complexity do separate jobs: quality governs visual detail and fidelity, while complexity governs how much goes into the scene. For most editing tasks, OpenAI recommends low complexity for faster edits, paired with whatever quality the workflow needs. Medium or high complexity is for scenes with many elements or constraints.

Two features carry over from ChatGPT or are new to the API. Image markup, already in ChatGPT, lets teams mark the exact area of an image they want changed. SVG exports produce vector files ready for websites, apps and print.

The 50% figure is OpenAI’s own, measured against “before”. When Images 2.5 launched on September 8, OpenAI said it had cut image generation latency by up to 50% compared with Images 2.0, as 9to5Mac reported then.

What the API costs

Testing Catalog lists Images 2.6 pricing for both text-to-image and image-to-image at $8/$5 for input and $30 for output, and says OpenAI’s pricing table carries both Flare and Sunburst. Those are the token rates the 2.5 models had: $5 per million text input tokens, $8 per million image input tokens and $30 per million image output tokens. With the 2.5 pair, identical published rates and Sunburst’s longer generation times meant there was no way to work out which model cost more per image, and the 2.6 rates arrive in the same form.

GIFs, sketches and saved references

GIFs are aimed at marketers and creators: an animated brand mascot, a product shot turned into a loop, reactions and memes, or animated storyboards to test character movement before video production. OpenAI says even a simple sketch can become an animated character for videos or games. Town, a startup building an AI work assistant, used Images 2.6 in the API to try animated characters in everyday work interactions.

The example prompts OpenArt wrote for the launch don’t ask for a GIF directly. Each asks for “a single animation sprite sheet for a seamless looping GIF”: exactly six frames in a 3-column by 2-row grid, every panel the same size, with the sixth frame transitioning back toward the first.

Reference images work as a saved library. OpenAI posted that users can keep reference images of themselves, pets and characters ready to reuse in new creations without uploading them again. The release also aims to keep characters, products and brand details consistent through successive rounds of editing. OpenAI has published a system card covering its safety evaluations and safeguards for Images 2.6.

People now create 4 billion images a week across ChatGPT, Codex and the API, more than half a billion a day, OpenAI said. At the Images 2.5 launch in September, the company put that figure at more than 3 billion a week.