OpenAI released GPT Image 2.5 on September 8, 2026 as ChatGPT Images 2.5, with Flare and Sunburst in the API.
No reference uploaded — GPT Image 2.5 will create a new image from your prompt.
Your generated image will appear here.
OpenAI shipped GPT Image 2.5 on September 8, 2026 in two API flavours — Flare and Sunburst. Both are wired into this page, and you never have to think about which endpoint you're hitting.
OpenAI positions Flare as the default for most applications, delivering higher-quality images than GPT Image 2 at up to 50% lower latency. It's the tier we preselect for social posts, thumbnails, and high-volume runs.
Sunburst trades speed for control across edits. Use it for production-ready campaign creative and polished product imagery where a small change must not disturb the rest of the frame.
The headline upgrade is multi-turn consistency: subjects, textures, and approved regions survive repeated edits far better, which is what kills most AI retouching workflows.
Drop in a face, a product, and a style frame together. Reference fidelity improved in 2.5, so recognizable subjects stay recognizable instead of drifting into a lookalike.
Early testers singled out text rendering and complex layouts — useful for posters, packaging, and ad frames where a headline has to be readable, not decorative.
Pick 1K for drafts and 4K for hero frames, across square, portrait, landscape, and 21:9 cinematic crops. Credits update live so the cost is never a surprise.
Most GPT Image 2.5 sites make you choose text-to-image or image-to-image first. Here you just write a prompt; uploading a reference is what turns a generation into an edit.
GPT Image 2.5 sits beside Nano Banana and Seedance on NanoBananaTool, so a still can go straight from prompt to edit to video without leaving the workspace.
What shipped, how the two tiers differ, and how this page works.