OpenAI's new state-of-the-art image model — faster generation, in-chat Sketch and Templates, and two API models (GPT-Image-2.5 Flare and Sunburst).
Last verified · 2026-09-08 · by Moe Ameen
ChatGPT Images 2.5 is OpenAI's image-generation and editing model, announced September 8, 2026 as its new state-of-the-art image model. Against the prior Images 2.0 it generates up to 50% faster and produces more natural lighting and richer textures, preserves subjects from a reference photo more faithfully, and edits more reliably across multiple turns of a conversation — you can make a selective change and keep the rest of the image intact, and the style holds steady over a longer back-and-forth.
Two features are new inside ChatGPT. Sketch lets you draw a rough reference directly in the chat, by typing @Sketch, and use that drawing to steer composition — a scribble carries layout intent that a paragraph of prompt often cannot. Templates give prebuilt starting points for common formats like flyers, posters, and product photos, so a non-designer gets a usable frame to build on.
For developers, OpenAI added two models to the Images API. GPT-Image-2.5 Flare is the default for most applications — creator and social content, product imagery, visual search, prototyping, and high-volume generation — carrying the quality gains at the lower latency. GPT-Image-2.5 Sunburst is built for premium workflows that need tighter control across edits, like campaign creative and polished product shots, and takes longer to generate. It rolled out to all ChatGPT, ChatGPT Work, and Codex users across desktop, mobile, and web; API pricing is token-based and per-image token usage is not clearly documented, so confirm current rates on OpenAI's pricing page.
The boundary worth planning around: this is an image maker, not a content workflow. It produces and edits one image in a chat. It does not caption for social, turn that image into a branded multi-format set, keep a persona's face consistent across a month of posts, or publish anywhere.
A ChatGPT Images 2.5 generation is a great starting asset and a finished post is a different thing — and the gap between them is exactly where [Kompozy](/) lives. Generate a hero image, product shot, or template-based graphic in ChatGPT (or at volume through the Flare API), then bring it into Kompozy as the base for a whole set: drop it into a brand-exact [Carousel](/glossary/hyperframes) built with HyperFrames, pull the strongest line into a Quote Graphic, wrap it into Photo Posts with per-platform copy written in your voice through the [Persona Brief](/glossary/persona-brief), and schedule the lot across nine platforms plus blog and email with autopilot and a review step. One image becomes a coordinated, on-brand drop instead of a single file you still have to caption and post by hand.
The sharper fit is consistency. ChatGPT Images 2.5 holds a subject steady within one conversation, but it does not lock a persona's face to reuse identically across weeks of content — and that cross-calendar sameness is what a recognizable brand identity actually needs. Kompozy's [Persona Photos](/glossary/persona-shorts) and Persona Tweets use Google Gemini face-lock to keep the same face on every render, and the AI Influencer persona pool drives recurring branded identity across formats. So the division of labor is clean: use ChatGPT Images 2.5 to make the best individual image, and use Kompozy to make it a consistent, published presence across every platform.
It is OpenAI's new image-generation and editing model, announced September 8, 2026. It generates up to 50% faster than Images 2.0 with more natural lighting and better subject preservation, edits reliably across a multi-turn chat, and adds Sketch (draw a reference) and Templates (starting points for flyers, posters, product photos). Developers use two API models, Flare and Sunburst.
Flare is the fast default for most uses — creator and social content, product imagery, prototyping, and high-volume generation. Sunburst is built for premium workflows needing tighter control across edits, like campaign creative and polished product shots, and takes longer to generate.
Generate the image in ChatGPT, then use a content engine to turn it into posts. On its own the model makes one image and stops. In Kompozy you fan that image into a branded carousel, quote graphic, and Photo Posts, add per-platform captions in your voice, and schedule them across nine platforms — the steps the model does not do.
It preserves a subject well within a single conversation, but it does not lock one persona identity to reuse identically across weeks of content. For a face that stays consistent on every post over time, a persona system like Kompozy's Gemini face-lock is the better fit.
It is available to all ChatGPT tiers, including free, for in-app use, with paid tiers getting faster and higher-volume image creation. The Images API is billed by token. Some plan-level image caps are not officially published, so confirm current limits and API pricing at OpenAI's pricing page.