OpenAI Unveils ChatGPT Images 2.5 With Boosted Speed and Fidelity

On Tuesday, OpenAI rolled out ChatGPT Images 2.5, an update that enhances image quality, editing accuracy, and the speed of its image creation tools.

According to OpenAI, users currently generate over 3 billion images weekly via ChatGPT Images and the GPT-Image models accessed through the API. The new version slashes generation time by up to 50% compared with Images 2.0.

OpenAI says Images 2.5 delivers more natural lighting and richer textures. It also preserves subject likeness from reference photos and follows editing instructions more consistently across multiple turns.

For ChatGPT users, the update introduces several features. Sketch allows users to draw directly in ChatGPT as a reference for the desired image. Templates provide starting points for common formats such as flyers and product photos. Users can now comment directly on images for focused editing and share prompts with others.

The rollout encompasses all ChatGPT, ChatGPT Work, and Codex users across desktop, mobile, and web platforms.

On the developer side, OpenAI released two new API models. GPT-Image-2.5 Flare mirrors the same quality, editing, and speed improvements, while GPT-Image-2.5 Sunburst offers added precision for detailed creative work but with longer generation times.

The new models also improve reference-photo handling, making subjects appear more recognizable while retaining their distinctive features across various settings and visual styles.

Images 2.5 excels at executing focused edits while preserving all other details, even with complex subjects and backgrounds. In longer conversations, the model maintains consistency across multiple edits without degrading image quality over time.

The update includes a better grasp of intricate visual instructions and more accurate interpretation of real-world information. The model can manage more complex layouts, including transparent backgrounds, and reflects visual styles more faithfully.

OpenAI stated that the release builds on existing safeguards with checks on prompts and images to prevent harmful outputs. The company continues using C2PA metadata and invisible watermarking so images created with its tools can be identified.

This article was prepared with the assistance of artificial intelligence and reviewed by an editor.