OpenAI shipped Images 2.6 across ChatGPT, Codex, and the API on November 4, and for developers the headline is a new model family split into two purpose-built variants rather than one general-purpose update.
Flare and Sunburst, two different jobs
GPT-Image-2.6 Flare is built for fast iteration and interactive workflows — OpenAI says it can return an image in as little as six seconds at low quality, which is squarely aimed at the back-and-forth of prompting and re-prompting rather than a single final render. GPT-Image-2.6 Sunburst takes the opposite priority: detailed compositions and consistency across edits at medium or high quality, for work where the first image matters more than the fifth. Framed together, it’s less “one model got better” and more “pick the tradeoff you actually need,” which is a shift from treating speed and fidelity as a single dial.
Across the family, OpenAI states Images 2.6 generates images 50% faster than the previous release — a number worth treating as OpenAI’s own claim rather than an independently benchmarked figure, since the company hasn’t published comparative benchmark scores or a per-image cost table for 2.6.
What’s new for ChatGPT users
The consumer side of the update turns a prompt or an uploaded photo into an animated GIF: users can adjust animation speed, edit individual frames, and share the result directly. That’s a meaningfully different capability than static image generation — it implies OpenAI’s pipeline is now tracking and editing per-frame state, not just producing a single flattened output.
SVG export and markup-guided edits
For anyone pulling generated images into a design or production pipeline, two additions matter more than the speed number: SVG export, which hands back vector output instead of a raster file, and markup-guided edits, which let a user annotate directly on an image to steer a targeted edit rather than re-describing the whole scene in a text prompt. A new complexity parameter gives direct control over how much detail a generation attempts, independent of the quality setting.
The competitive backdrop
Images 2.6 lands into a crowded field: Google’s Gemini API already ships its own “Nano Banana” image line, and Black Forest Labs continues iterating FLUX on its own release cadence. GIF output and SVG export are OpenAI’s attempt to differentiate on practical, pipeline-friendly formats rather than purely on image quality — a reasonable bet given how much generative-image competition has become a benchmark race that most working artists and designers never look at directly.