ChatGPT Images 2.5 release notes / what's new

OpenAI has launched ChatGPT Images 2.5, a state-of-the-art image model designed to improve visual fidelity, editing precision, and generation speed. The update introduces a 50% reduction in image generation latency compared to Images 2.0 and adds new interactive tools for users to guide the creative process.

Improved Image Fidelity and Reference Handling

Images 2.5 enhances the ability to transform subjects from reference photos across different settings, styles, and compositions while maintaining the subject's recognizability. The model produces more natural lighting, richer textures, and better preservation of distinctive features, making reference-led workflows more reliable for both end-users and API developers.

Precision and Multi-Turn Editing

Editing in Images 2.5 is more focused, allowing users to modify specific elements—such as a product or background—without altering the rest of the image. This precision is maintained across multiple conversational turns, ensuring that earlier changes remain consistent and image quality does not degrade as instructions become more specific.

Intelligence, Style, and Layout

The model demonstrates improved understanding of complex visual instructions and real-world information. Key technical improvements include:

  • Complex Layouts: Better handling of layouts, including the generation of transparent backgrounds.
  • Visual Style Alignment: Increased accuracy in reflecting requested artistic visions and creative briefs.
  • Content Accuracy: More accurate representation of real-world information within images.

New Creative Tools in ChatGPT

OpenAI has introduced several product features to give users more control over image creation:

  • Sketch: A feature allowing users to draw rough layouts or doodles directly in ChatGPT (via the "@Sketch" command) to serve as a visual guide for the final image.
  • Templates: Pre-defined formats for popular creative needs, such as "Poster" or "Merch," to help users structure their prompts.
  • Direct Comments: The ability to place comments directly on images for more focused editing.
  • Prompt Sharing: Users can now share the specific prompts used to generate an image, allowing others to replicate or adapt the ideas.

API Model Options

For developers, OpenAI is releasing two distinct models via the API:

  • GPT-Image-2.5 Flare: The default choice for most applications, offering the same quality and speed improvements as the chat interface, including 50% lower latency than GPT-Image-2.
  • GPT-Image-2.5 Sunburst: A model optimized for premium visual workflows requiring higher precision and tighter control across edits, though it involves longer generation times.

Safety and Availability

ChatGPT Images 2.5 is available to all ChatGPT, ChatGPT Work, and Codex users across desktop, mobile, and web. To ensure safety, the model employs prompt and image checks to prevent harmful outputs, utilizes C2PA metadata, and implements invisible watermarking. Detailed evaluations are available in the system card.

Sources