ChatGPT Images 2.5 is rolling out across ChatGPT, ChatGPT Work and Codex with faster generation, more controlled multi-turn edits, a mobile Sketch tool and two new API models. OpenAI says in its launch post that the release cuts generation latency by as much as 50% against Images 2.0, but that performance figure is the company’s own measurement rather than an independent benchmark.
Everyone else is reporting a faster image model; we are explaining why the workflow features around the model may matter more to production teams than a headline speed claim.
What ChatGPT Images 2.5 actually changes
OpenAI’s September 8 product post says ChatGPT Images 2.5 improves natural lighting, texture, reference-subject preservation and compliance with editing instructions. The company is also separating its developer offering into GPT-Image-2.5 Flare, designed for faster work, and GPT-Image-2.5 Sunburst, intended for more exacting edits that can tolerate longer waits.
Axios received early access and described the launch as both an image-engine upgrade and a redesigned editor. Its hands-on test covered tasks including adapting a photograph, making a logo and developing a tattoo concept. TechRadar’s early assessment similarly focused on practical controls rather than raw model scores, highlighting Sketch, targeted comments and iteration as the most immediately useful changes.
| Item | Verified detail |
|---|---|
| Launch date | September 8, 2026 |
| ChatGPT availability | All tiers on desktop, mobile and web |
| API models | GPT-Image-2.5 Flare and GPT-Image-2.5 Sunburst |
| New controls | Sketch, templates, comments and prompt sharing |
| Company speed claim | Up to 50% lower latency than Images 2.0 |
ChatGPT Images 2.5 moves editing closer to a creative loop
The more consequential change is persistence across edits. Image generators can make a strong first frame and then lose a face, product detail or composition as a user asks for small adjustments. OpenAI says the new model is more likely to preserve earlier choices through a longer conversation. If that holds up in independent use, it reduces the need to restart from the original prompt every time one element changes.
Sketch gives mobile users a second input channel: draw a rough shape or arrangement, then describe what the final image should become. Comments let a user point to the part that needs work. Templates shorten the blank-canvas stage, while prompt sharing turns a successful recipe into something another person can reuse with their own material.
Why the two-model API split matters
Production teams rarely want one universal setting. A commerce catalogue may value throughput, while a campaign hero requires careful typography, layout and repeated corrections. Flare and Sunburst make that trade-off explicit at model selection rather than hiding it behind one interface. Developers can route routine variants to the faster option and reserve the slower model for assets where precision is worth the additional time.
The release also lands as companies connect image systems to broader agent workflows. Lapaas Voice recently examined how Motif Design is embedding agents in browser-based design and how the Logitech MX Keypad gives creators physical AI controls. ChatGPT Images 2.5 fits that pattern: the value is increasingly in how a model accepts context, preserves state and slots into a repeatable process.
Safety and provenance remain part of the release
OpenAI’s system card says the higher realism of the new models can increase the risk of convincing deepfakes. The company says it uses prompt- and image-layer safeguards, C2PA metadata and invisible watermarking. Those measures do not eliminate the need for review: teams still have to verify likeness rights, product accuracy, text and source provenance before publishing generated work.
In short: ChatGPT Images 2.5 is less about a single spectacular output than a shorter path from rough idea to controlled revision. The independent tests available on launch day are early, so claims about speed and fidelity should be treated as provisional until users test them across varied workloads.
Frequently asked questions
What is ChatGPT Images 2.5?
It is OpenAI’s updated image-generation and editing system for ChatGPT, ChatGPT Work, Codex and the API, with new workflow controls and two developer models.
What is the difference between Flare and Sunburst?
OpenAI positions Flare as the faster API model and Sunburst as the higher-precision option for detailed creative work where longer generation time is acceptable.
Is ChatGPT Images 2.5 available now?
OpenAI says it began rolling out on September 8 across all ChatGPT tiers and made both API models available the same day.
Get the day’s top stories in your inbox
One concise email. No spam, unsubscribe anytime.
