OpenAI Ships ChatGPT Images 2.5 With Sketch Tool and 50% Faster Generation

OpenAI says people now create more than 3 billion images every week across ChatGPT Images and the GPT-Image models in the API. Today the company launched ChatGPT Images 2.5, a new image model it calls state-of-the-art, promising sharper details, more precise editing, and faster generation. The release also introduces Sketch, a drawing feature inside ChatGPT, alongside templates, on-image comments, and shareable prompts.

The scale matters. Image generation has become one of OpenAI's highest-volume consumer surfaces, and the company is competing with Adobe, Google, and a wave of startups for creative workflows that span marketing, retail, and media production. Images 2.5 is OpenAI's bid to move from novelty generation toward dependable production tooling.

What changed under the hood

Images 2.5 reduces image generation latency by up to 50% compared with Images 2.0, according to OpenAI. The model produces more natural lighting and richer textures, preserves subjects from reference photos more faithfully, and follows editing instructions more reliably across multiple turns of a conversation.

Reference fidelity is the headline improvement. OpenAI says images built from reference photos now keep subjects more recognizable, with distinctive features more likely to carry through as users move familiar people and places across new settings, visual styles, and compositions. For teams building on the API, that reliability lets variations stay anchored to an original source image.

Editing precision improved as well. The model edits only what users ask for while keeping surrounding details intact, even with complex subjects and backgrounds. In the Images API, that means developers can update a single element — a product, a background, or a piece of copy — while preserving the subject, composition, and brand treatment around it.

Multi-turn consistency also got attention. OpenAI says earlier changes now stay consistent across successive edits, with each new edit building on previous work without degrading image quality. The company notes this matters for production workflows where developers need targeted changes without rebuilding an entire asset.

On intelligence and style, Images 2.5 handles complex visual instructions better, produces more accurate content in images containing real-world information, supports complex layouts including transparent backgrounds, and reflects requested visual styles with less drift as instructions become more specific. OpenAI frames this as making complex creative briefs more dependable for businesses generating on-brand asset series, UI concepts that preserve a hierarchy, or presentation visuals that fit a defined structure.

Sketch: draw your way to an image

The most notable new product feature is Sketch. Users can draw directly in ChatGPT — the layout of a room, the contour of an outfit, or, in OpenAI's words, "just a funny doodle" — and ChatGPT turns the rough art into a complete image. Users add a description of the style and other details to steer the result. Access requires typing "@Sketch" in ChatGPT.

"You don't need to be a professional artist — this is just another way to get your final image closer to what you have in mind," OpenAI writes.

Templates arrive alongside Sketch for popular creative formats. Instead of a blank canvas, users can pick a template like "Poster" or "Merch" and add information, design elements, or styles to personalize the output.

Sharing changed too. When users share an image, they can now include the prompt that produced it, letting others run the same idea with their own images and details. OpenAI points to a currently viral prompt that renders what users would have looked like in the 1980s. Users can also place comments directly on images for more focused editing.

Two new API models

For developers, OpenAI released two models in the API. GPT-Image-2.5 Flare brings the same quality, editing, and speed improvements to the API and serves as the default for most applications, delivering higher-quality images than GPT-Image-2 at 50% lower latency. OpenAI positions Flare for creator and social content, product experiences, visual search, rapid prototyping, and high-volume generation.

GPT-Image-2.5 Sunburst targets premium visual workflows that need tighter control across edits, at the cost of longer generation times. OpenAI suggests it for production-ready campaign creative and polished product imagery. The company lists early customer feedback on its announcement page but did not publish pricing figures in the announcement itself; pricing details are available on the OpenAI developers site.

The two-tier split mirrors a broader industry pattern: fast, cheap generation for volume use cases and slower, higher-precision models for finished commercial work.

Safety and availability

Images 2.5 builds on OpenAI's existing safeguards, with checks on both prompts and images to help prevent harmful outputs. The company continues to apply C2PA metadata and invisible watermarking to identify images made with its tools. OpenAI published a system card detailing its evaluations and safety approach.

Availability is broad from day one. Images 2.5 is rolling out to all ChatGPT, ChatGPT Work, and Codex users across all tiers on desktop, mobile, and web. GPT-Image-2.5 Flare and Sunburst are available in the API now.

With 3 billion images a week already flowing through its image stack, OpenAI's challenge with Images 2.5 is less about adoption than whether reference fidelity and multi-turn consistency hold up in the production workflows where Adobe and Google currently dominate.