If you've ever generated an image in ChatGPT, waited ten seconds, then watched it botch the face you were trying to keep consistent across edits, OpenAI's newest release is aimed squarely at you. On September 8, 2026, OpenAI began rolling out ChatGPT Images 2.5 to all ChatGPT, ChatGPT Work, and Codex users across desktop, mobile, and web — alongside two new image models in the API, GPT-Image-2.5 Flare and GPT-Image-2.5 Sunburst. For agencies and businesses that have started weaving AI-generated visuals into real client work, this update matters less as a headline feature and more as a quiet fix to the things that made AI image tools unreliable for production use.
The topline number is speed: OpenAI says the new model generates images up to 50% faster than its predecessor, which turns image generation from a "start it and go get coffee" task into something closer to real-time iteration. But the more consequential improvements are about fidelity and consistency. Images 2.5 produces more natural lighting and richer textures, and it's noticeably better at preserving the actual identity of people and objects in reference photos — a persistent weak spot for generative image tools, where a subject's face or a product's exact shape would subtly drift with every edit.
OpenAI also says the model follows editing instructions more reliably across multiple turns, which addresses a common failure mode: ask for one small change on the third edit and the model quietly forgets constraints you set on the first. A new Sketch feature lets users draw directly inside ChatGPT as a visual reference for the image they want generated, and OpenAI has added comment-based editing, so you can annotate specific parts of an image with notes rather than describing changes in a single paragraph of prompt text.
On the API side, GPT-Image-2.5 Flare carries the same quality, editing, and speed gains into programmatic workflows, while GPT-Image-2.5 Sunburst trades speed for extra precision on detailed creative work — useful for anyone building an automated pipeline where a slower, more accurate pass is worth the wait.
Most AI-native agencies and no-code builders have already tried routing image generation into their workflows — product mockups, ad creative variations, social content, quick client concepts — and hit the same wall: outputs looked good in isolation but fell apart across a sequence of edits, or took long enough to generate that they broke the flow of an automated pipeline. Faster generation plus better multi-turn consistency directly addresses both of those failure points.
The Sketch feature is also a genuinely practical addition for client work. Instead of trying to describe a layout in words and hoping the model interprets it correctly, a designer or account manager can rough out a composition by hand and let the model fill in the rest — closer to how creative direction actually happens in an agency setting.
If you already have ChatGPT or ChatGPT Work access, Images 2.5 should already be live in your account — there's no separate opt-in. For teams building automated content pipelines (say, generating on-brand social assets from a content calendar, or producing first-pass product photography variants), the two new API models are the more relevant entry point:
Either way, the API access means these improvements can be dropped into existing n8n, Zapier, or custom automation flows without waiting on a UI redesign.
RAG assistants on your documents, AI agents that act in your tools, and LLM features inside your product.
Our cutting-edge features simplify collaboration and creativity, making your workflow intuitive and efficient. Transform your vision into reality effortlessly with Hadidiz Flow.
