OpenAI ChatGPT Images 2.6 Guide

ChatGPT Images 2.6 is OpenAI’s newest image model in ChatGPT and Codex, rolled out from November 4, 2026. It follows Images 2.5 and adds animated GIFs, saved reference images, faster generation, and more precise editing.

Images 2.6 is less about one perfect picture and more about keeping a whole project consistent.

• Rolling out to every ChatGPT plan, including free
• On web, desktop, and mobile
• In ChatGPT and Codex
• Developers get two new API models

Usage limits differ by plan. Check ChatGPT’s help pages for current limits.

GIFs

Turn characters, mascots, product shots, sketches, and storyboards into short animated loops

Saved references

Save reference images of yourself, pets, and characters and reuse them without re-uploading

Speed

OpenAI says images generate about 50% faster than before

Consistency

Characters, products, and brand details stay more consistent across rounds of edits

Turn this sketch of our coffee cup mascot into a short looping GIF: the mascot waves, steam rises from its head, simple flat colours, plain background.

• Describe one simple, repeatable motion
• Keep the background plain for clean loops
• Ask for a seamless loop

• Save a clear reference of a character, pet, or product
• Name it so you can call it in later prompts
• Use only images of people who have agreed, or of yourself

• Use image markup to point at the exact area to change
• Make one change per request
• Say what must stay the same: face, logo, colours, layout

Change only the jacket to dark green. Keep the face, pose, background, and lighting exactly the same.

• Subject, setting, style, lighting, and framing
• Exact text in quotes for signs and posters
• The format: poster, product photo, icon, or storyboard
• The aspect ratio you need

GPT-Image-2.6 Flare

Built for speed and professional layouts such as posters and product images; as little as six seconds at low quality

GPT-Image-2.6 Sunburst

Built for detail and consistent edits at medium or high quality

• Image markup is now in the API
• A complexity setting controls how much planning the model does: low for quick edits, medium or high for busy scenes
• SVG export for websites, apps, and print
• Priced per token like Images 2.5; check OpenAI’s pricing page

• Social graphics and animated reactions
• Product mockups and ad variations
• Storyboards and character sheets
• Icons and logos as SVG
• Diagrams and slides

• Do not create misleading images of real people or events
• Respect trademarks and copyrighted characters
• Label AI images where platforms or laws require it
• Check text and details before publishing