OpenAI’s Images Just Got Better: New Features and 50% Improved Latency
OpenAI has introduced ChatGPT Images 2.5, a new image model for creating and editing images in ChatGPT and through the Images API. The update focuses on sharper details, more reliable editing, improved style and instruction following, and faster generation, with latency reduced by up to 50% compared with Images 2.0.
Images 2.5 is designed to preserve the identity and details of subjects taken from reference photos. OpenAI says the model produces more natural lighting and richer textures, while distinctive features are more likely to carry through when a subject is placed in a different setting, visual style, or composition. This also supports reference-led workflows in the API, where variations need to remain anchored to the original source.
Editing is another central part of the update. The model is better at changing only the requested element while keeping other details intact, including complex subjects and backgrounds. In API workflows, that can mean updating a product, background, or piece of copy without rebuilding the surrounding composition or brand treatment.
The model also aims to maintain consistency across multiple editing turns. Earlier changes are more likely to remain stable as new instructions are added, and each edit is intended to build on the previous work without gradually reducing image quality. Images 2.5 is also better at interpreting complex visual instructions, handling layouts with transparent backgrounds, and reflecting requested visual styles. OpenAI says real-world information in images is more accurate and that the model is more likely to retain specific composition and design details as a brief becomes more detailed.
ChatGPT is gaining several tools alongside the model. Sketch lets users draw directly in ChatGPT and use the drawing as a visual guide for a final image. Templates provide starting points for formats such as posters and merchandise, while comments can be placed directly on images to support more focused editing. Users can also share the prompts behind images, allowing others to reuse the underlying idea with their own images and details.
For developers, OpenAI is releasing GPT-Image-2.5 Flare and GPT-Image-2.5 Sunburst in the API. Flare provides the quality, editing improvements, and faster generation of the new release, with 50% lower latency than GPT-Image-2, and is positioned as the default option for most applications. Sunburst is intended for workflows that require additional precision and can accommodate longer generation times.
Images 2.5 is rolling out to ChatGPT, ChatGPT Work, and Codex users across all tiers on desktop, mobile, and web. The update combines new creation and editing tools with improved fidelity, multi-turn consistency, and latency reduced by up to 50%.
Sources:
