Summary
OpenAI has introduced Images 2.6 in ChatGPT and Codex, and made two model variants available through its API. The release adds GIF creation, image-markup editing in the API and SVG exports, alongside company-reported speed and consistency improvements.
OpenAI has introduced Images 2.6 in ChatGPT and Codex, with two versions for API users: GPT-Image-2.6 Flare and GPT-Image-2.6 Sunburst. The release adds GIF creation and new editing options for developers, including image markup and SVG exports. OpenAI says the model generates images 50% faster than its previous offering; the announcement does not specify the comparison model or test conditions behind that figure.
Faster generation, with separate controls for quality and complexity
The two API variants are aimed at different needs. OpenAI describes Flare as the faster option and says that, with quality set to low, it can generate images in as little as six seconds. It also says Flare improves layout quality for posters, product imagery and other professional visuals. Sunburst is intended for work requiring more detail and precision, with OpenAI highlighting image consistency across edits at medium or high quality settings.
The API also gains a complexity parameter. Developers can use it to adjust how much planning the model applies to a request: OpenAI recommends lower complexity for quick edits and medium or high settings for scenes with many elements or constraints. Image quality is a separate control for visual detail and fidelity. In practical terms, these settings let a developer make a trade-off between the effort spent following a complicated request and the level of visual detail requested.
Image markup, previously available in ChatGPT, is now available through the API. It lets a team identify a specific area for a requested change rather than describing the whole image again. SVG exports are also supported. SVG is a vector format, which can be scaled for different sizes without the pixel-based loss of detail associated with enlarging a raster image; OpenAI says the exports are intended for use in websites, apps and print.
GIFs add motion to image workflows
Images 2.6 can be used to create GIFs, bringing short looping animations into the image-generation workflow. OpenAI presents examples such as animated brand mascots, social-media reactions and storyboards for exploring character movement before video production. The announcement also describes starting from a simple sketch to develop an animated character.
These uses make consistency across editing rounds relevant: a character or product needs to retain defining features as its pose or surrounding details change. OpenAI says Images 2.6 is designed to preserve characters, products and brand details during edits, and highlights GIF creation as a way to extend static visuals into motion. The announcement includes examples from OpenArt and describes API use by Town, an AI work-assistant startup.
OpenAI says Images 2.6 is available globally in ChatGPT and Codex across all subscription tiers on desktop, mobile and web. GPT-Image-2.6 Flare and Sunburst are available in the API.