OpenAI's image API can now output SVG and animated GIFs directly
GPT Image 2.6 adds two models, Flare and Sunburst, plus markup-guided editing. SVG output comes with real restrictions: one image per request, no streaming, no edits.
OpenAI shipped Images 2.6 on 4 November, rolling out two new models in its Image API: Flare, which generates images in as little as six seconds, and Sunburst, aimed at higher quality and more consistent edits. OpenAI says overall generation is 50% faster across the update.
The headline addition for builders is direct SVG output from the generation endpoint, vector markup rather than a raster file you’d have to trace yourself. It comes with real limits: SVG requests must use n=1, can’t stream, and don’t support partial images, prompt_rewriter or output_compression. There’s no SVG editing, only generation. The edits endpoint separately gains markup-guided editing: you send a transparent “markup” layer instead of a mask, and the transparent areas mark where edits are allowed. The API can also produce animated GIFs now, with per-frame editing and adjustable speed.
Why it matters: scalable vector output straight from a prompt is new for a flagship image API, useful for icons, diagrams or illustrations you’d otherwise need a designer or a manual trace step to vectorise. Markup-guided editing is also a cleaner primitive than painting a mask if you’re building an image-editing feature on top of the API.
The caveat: SVG output is billed at the image model’s rate for the reference render plus the authoring model’s token rate for producing the SVG markup, so a vector asset can cost more than the raster image it’s derived from. Check the pricing for your actual use case before building on it.