GPT Images 2.6 Operational

GPT Images 2.6 is OpenAI's next-generation image generation and editing model, now fully available across ChatGPT, Codex, and the API. It generates images 50% faster, with results in as little as 6 seconds; on the API, it is split into two models, Flare and Sunburst. The complexity and quality parameters separately control scene planning and visual detail, balancing speed and consistency. GPT Images 2.6 natively supports GIF generation and SVG vector export, and brings region-based inpainting to the API.

Faster generation: Overall generation time is reduced by 50%. The Flare model can produce an image in as little as 6 seconds at low quality, with layout optimization for commercial visuals such as posters and e-commerce product images.
Two-model division of labor: The API is split into the Flare and Sunburst models.
Native GIF generation: Generate clean-background, fluid looping GIFs directly for brand mascots, stickers, and subtle facial-expression animations for virtual assistants.
SVG vector export: Export natively in SVG format. Icons and illustrations can be scaled infinitely, eliminating third-party raster-to-vector conversion and allowing direct insertion into frontend codebases.
API region-based inpainting: The region-selection editing feature previously available only on the ChatGPT web interface is now available through the API. Specify edit areas with coordinates or masks and process image edits in batch pipelines.

Decoupled planning and rendering: The API adds a complexity parameter that controls how much compute the model spends planning a scene before drawing. The existing quality parameter focuses solely on visual detail and fidelity. For routine local adjustments, lower complexity while keeping quality high. The model can then skip a complete reconstruction of the composition and perform only high-precision inpainting. This is similar to adjusting reasoning depth in large language models and can significantly reduce the cost and time of multi-round edits.
GIF frame-sequence approach: The model borrows the game industry's Sprite Sheet concept, drawing a 3-column × 2-row sequence of 6 animation frames in a single image. Each frame must have the same dimensions and fixed camera position, while the character and scene elements remain unchanged and only expressions and movements progress frame by frame. The final frame transitions smoothly back to the first, after which the frames are sliced and combined into a GIF.
Sunburst's consistency mechanism: Sunburst provides greater detail accuracy at medium and high quality levels. It primarily addresses the problem of people and products gradually changing during multi-turn edits. Combined with the API's image annotation capabilities, it enables batch edits to iterate reliably from a consistent visual baseline.

Access: GPT Images 2.6 is fully available across ChatGPT, Codex, and the API. Free and paid users can use it directly on desktop, mobile, or the web.
Text-to-image: Describe your needs in natural language in ChatGPT or Codex to generate images. For faster generation, choose the speed-focused Flare model.
Parameter controls: When calling the API, use the complexity parameter to control scene-planning compute and the quality parameter to control visual detail and fidelity.
Generate GIFs: Ask for a seamless looping GIF and describe the action sequence and fixed camera position in the prompt. The model will automatically produce an animation image arranged as sequential frames for slicing and GIF assembly.
Export SVG: When you need vector graphics, request SVG output directly. Icons, components, and illustrations can be scaled infinitely and inserted directly into a frontend codebase.
Local edits: In the ChatGPT web interface or through the API, use region selection together with coordinates or masks to specify edit areas. For multi-round edits, the more consistent Sunburst model is recommended.

Twice the speed: Overall generation time is reduced by 50%, and the Flare model can produce an image in as little as 6 seconds at low quality, enabling low-cost rapid iteration.
Strong controllability: The complexity and quality parameters decouple scene planning from visual rendering, allowing developers to fine-tune each independently.
High consistency: The Sunburst model keeps people, products, and brand details stable during multi-turn edits, reducing unwanted changes over time.
Native animation: Generate smooth looping GIFs directly through a frame-sequence approach without relying on a video model, with low latency and fixed elements between frames.
Engineering-friendly: Native SVG vector export and API region-based editing eliminate format conversion and can be embedded directly into frontend and batch image-editing workflows.

Official website: https://openai.com/index/introducing-images-2-6/

E-commerce and brand marketing: Quickly generate posters and product images with the Flare model. With images produced in 6 seconds at low quality, it is well suited to high-frequency commercial visual iteration.
Social media content: Generate looping GIFs natively for brand mascot animations, stickers, and reaction graphics that fit social sharing contexts.
Video pre-production: Explore character movements and shot pacing with animated storyboards before investing in full video generation.
Frontend and UI development: Generate SVG icons, components, and illustrations directly in Codex and add them to a codebase, eliminating the intermediate raster-to-vector conversion step.
Embedded AI interactions: Use the API to generate real-time micro-expression animations such as typing and feedback states for virtual assistants and other products, or use image annotation capabilities to batch-edit images and build enterprise-grade retouching pipelines.