OpenAI has launched a new image generation model called ChatGPT Images 2.5. This model produces significantly sharper AI images and delivers results faster than before. Editing tools have also been improved. The company states that users generate over three billion images weekly using ChatGPT Images and GPT-Image models, a process that also takes place via the API. The company now aims to make this process even easier.
Availability and Speed
Images 2.5 is now available to all ChatGPT users, across both free and paid versions. Users of ChatGPT Work and Codex can also access it. The feature has rolled out across desktop, mobile, and web platforms. GPT-Image-2.5 Flare and GPT-Image-2.5 Sunburst are immediately available via the API.
According to OpenAI, lighting in Images 2.5 appears more natural compared to previous versions, and textures look better. When using a reference photo, the model preserves the original subject accurately. It also follows editing instructions more effectively. Compared to the older model (Images 2.0), Images 2.5 reduces image generation time by up to 50%.
New Features
OpenAI has introduced several new tools, most notably the ‘Sketch’ tool. This allows users to draw directly within ChatGPT—creating anything from room layouts to simple doodles—and ChatGPT will generate a full image based on that sketch. Additionally, templates for flyers and product photos are provided. Users can request specific changes by adding comments to the image. Prompts used for generation can also be shared.
The new model offers improved ‘focused editing’; it modifies only the specific area requested by the user while keeping the rest of the image intact. Images also remain more consistent throughout long conversations compared to previous versions. The new model accurately interprets complex instructions, detailed layouts, and transparent backgrounds.
Models for Developers
Two models have been released for developers. GPT-Image 2.5 Flare offers improvements in both quality and speed; it is ideal for creating social media content, product images, and prototypes. GPT-Image 2.5 Sunburst is designed for more detailed work; it offers higher precision but requires slightly more time. It is tailored for tasks such as advertising and polished product photography.
