OpenAI launches ChatGPT Images 2.5 with faster editing
OpenAI has launched ChatGPT Images 2.5, the latest iteration of its in-product image generation system. The update centers on speed and editing precision, with the company saying that generation times are noticeably reduced and that multi-turn edits - where users refine an image across several back-and-forth instructions - now produce more accurate results than the previous version.
Two headline features arrive alongside the core improvements. Sketch allows users to provide a rough drawn input that the model then develops into a more polished image, giving artists and designers a way to guide outputs with a visual starting point rather than text alone. Template support, meanwhile, lets users apply consistent structural or stylistic frameworks across images, which is useful for anyone producing recurring content formats or branded material.
On the developer side, OpenAI is introducing two new API models tied to this release. This expands the options available to teams building image generation into their own products, and suggests OpenAI is treating the image API as a meaningful product line rather than a secondary offering. The previous image generation API, which exposed the model behind the in-app experience, had already seen adoption across a range of third-party tools.
ChatGPT Images 2.5 continues a period of steady investment by OpenAI in native image capabilities within ChatGPT. The broader context is a competitive one - Adobe Firefly, Google Imagen, and a range of independent tools are all vying for both consumer and developer attention in the generative image space. Faster editing and stronger multi-turn coherence are practical improvements that address common friction points users encounter when iterating on generated images, and the addition of Sketch and templates signals a push toward more structured, workflow-oriented use cases.
