OpenAI unveils ChatGPT Images 2.5, boosting speed, detail, and collaborative features across platforms, posing new challenges for creative workflows and proprietary control.
OpenAI has updated its image tools with ChatGPT Images 2.5, a release that emphasises faster generation, sharper detail and more controlled editing rather than a major shift in model architecture. The company says the system can now produce images with up to 50% lower latency than Images 2.0, while also improving fidelity when users work from reference photos. OpenAI also said the update is available across ChatGPT, ChatGPT Work and Codex on desktop, mobile and the web.
The most striking part of the release is how much of it is about the interface. Users can now sketch directly inside ChatGPT, place comments on images and start from templates for common formats such as posters and merchandise. OpenAI has also added the option to share the prompt alongside an image, allowing other people to reuse the same instruction with their own photos. That turns prompts into portable creative assets, which is as much a distribution strategy as a convenience feature.
The API side now includes two new models: GPT-Image-2.5 Flare, which OpenAI positions as the default option, and Sunburst, which is aimed at higher-control production work. Adobe has already moved quickly to support the update in Firefly, its creative AI studio. Adobe says the newer GPT image model is now available there, with the company presenting Firefly as a place where users can choose between models while staying inside Adobe’s workflow.
That matters because the competition in creative software is increasingly about where work begins. OpenAI and Adobe are each becoming a channel for the other: Adobe tools are available inside ChatGPT, while OpenAI’s image models are now part of Firefly. The result is a contest over the first step in a creative task, not just the quality of the underlying model. In that market, whichever platform captures the opening prompt has the advantage.
OpenAI said more than 3 billion images are created each week across ChatGPT Images and its image API models. That figure is the company’s own and is not independently verified, but it points to the scale at which synthetic imagery is now being generated. OpenAI says it is attaching C2PA metadata and invisible watermarks, alongside prompt and image checks, and has published a system card for the release. At this volume, provenance tools are no longer optional extras; they are part of the core infrastructure.
There are still limits to that protection. Metadata can be stripped when images are reposted, screenshotted or re-encoded, and watermarking systems vary in robustness. OpenAI’s prompt-sharing feature also cuts both ways: it makes image creation easier to repeat, but it can also help spread a single visual style very quickly. For Adobe, the strategic question is whether Firefly can remain a broad layer over multiple models, rather than become dependent on one supplier.
Disclaimer: This content is intended for informational purposes only. Readers are advised to exercise their own judgement, conduct due diligence, or consult a qualified expert before acting on any information provided.
