GPT Image 2.6

OpenAI's latest image model, released on 4 November 2026 in ChatGPT, Codex and the API. It updates GPT Image 2.5, and it is the first version that can return an animated GIF besides still images.

They are versions of the same model family. GPT Image 2.5 (September 2026) came in two variants: Flare, the fast default, and Sunburst, for precise edits. GPT Image 2.6 keeps both (gpt-image-2.6-flare and gpt-image-2.6-sunburst) and the per-token prices of 2.5, and adds GIF output, a complexity setting, edits marked on an annotated image and SVG output.

Yes. OpenAI added GIF output with Images 2.6 on 4 November 2026, in ChatGPT and in the API. In Vuela you use the same model with styles tuned for loops, flicker removal on drawn styles and a library where every GIF is saved.

Up to 16 frames. At Vuela's normal speed (120 ms a frame) the loop lasts about 2 seconds; slow is 200 ms a frame and fast 80 ms. GIFs have no sound and use up to 256 colours.

Through OpenAI's API, a 16-frame GIF at 480x848 in low quality used 1,290 image tokens, about $0.039; high quality uses about twice as many, and bigger frames cost more. In Vuela a GIF costs 6 credits in standard quality and 12 in premium.

The model draws every frame again, so even where nothing moves about half of the pixels change slightly from one frame to the next. Vuela removes that flicker on pixel art, cartoon and sticker GIFs at no cost. On photorealistic GIFs the same fix would blotch skin and fabric, so they keep a slight flicker.

OpenAI doesn't document it. In Vuela every GIF comes out on a solid background: GIF transparency is all or nothing per pixel, which leaves rough edges around the subject.

Open the GIF tab of the image playground, describe the scene and the motion or start from an image, and pick a style, a format (square, 9:16, 16:9 or 4:5), a speed and a quality. Every image in your library also has a Convert to GIF button, and you can make GIFs from an AI assistant connected to Vuela's MCP server.