OpenAI Launches ChatGPT Images 2.5 with New API Models and Sketch Features

OpenAI released ChatGPT Images 2.5 on September 8, 2026, describing the release as its new state-of-the-art image model that processes over three billion images weekly across the API and ChatGPT. [3]

The updated model reduces image generation latency by up to 50 percent when compared directly with the previous Images 2.0 model. [3][4][5] According to OpenAI, Images 2.5 generates images with more natural lighting and richer textures than its predecessor. [3][5] Furthermore, the updated system is better at preserving subjects from reference photos and follows editing instructions more reliably across multiple turns of a conversation. [1][3] During these longer conversations, earlier edits are more likely to carry through and each new instruction builds on previous work without degrading the overall image quality. [3]

Alongside the core model upgrade, OpenAI introduced a new feature called Sketch, allowing users to draw ideas directly inside ChatGPT to serve as a visual reference for the final AI-generated image. [3][5] Users can invoke this specific drawing tool simply by typing "@Sketch" in a chat interface. [3]

The release also significantly expands inline image editing capabilities by allowing users to leave comments or annotations directly on the generated images for more focused editing. [3][5] Instead of starting entirely from scratch, users can now access ready-made templates for popular formats, with named starting points including "Poster" and "Merch." [3][5] OpenAI additionally noted that the model is better at modifying only the specific elements a user asks it to edit while leaving the rest of the complex background intact. [3]

For API developers, OpenAI released two distinct new image models to handle different generation requirements. [3][5] The first model, named GPT-Image-2.5 Flare, is designed specifically for fast, high-quality everyday image generation targeting social media content and product imagery. [1][5]

The second model, GPT-Image-2.5 Sunburst, is designed for creative tasks where precise control matters most, though its generation time may be longer compared to the Flare model. [1][5] The broader Images 2.5 update is currently rolling out to all users of ChatGPT, ChatGPT Work, and Codex across desktop, mobile, and web environments. [3][5]

OpenAI's latest release emphasizes operational speed and workflow retention, cutting generation latency by up to 50 percent compared to the previous Images 2.0 model. By introducing dedicated models for speed (Flare) versus precision (Sunburst), the company is actively segmenting its API offerings to better match distinct developer use cases and volume requirements. The addition of inline sketching, annotations, and template features directly addresses the historical challenge of maintaining consistent subjects and complex layouts across multiple conversational turns. What the sources don't address: the exact pricing structures for the newly introduced Flare and Sunburst API models.