OpenAI releases ChatGPT Images 2.5 with up to 50% lower latency and Sketch input
On September 9, OpenAI released ChatGPT Images 2.5 with up to 50% lower latency, improved reference fidelity, new Sketch input and templates, and two API variants Flare and Sunburst.
On September 9, OpenAI released ChatGPT Images 2.5. The new image model cuts generation latency by up to 50% and improves detail, lighting, texture rendering, and multi-turn editing stability. It is available to all ChatGPT, ChatGPT Work, and Codex users across desktop, mobile, and web.
Reference fidelity is improved: the model preserves subject features across new contexts, styles, and compositions. Local edits are more precise, leaving product, character, and overall composition intact while changing only what the user asks for. ChatGPT Images and the GPT-Image API together produce more than 3 billion images per week.
New product features include Sketch, which lets users sketch a rough composition as a visual guide; Templates for common formats such as flyers and product photos; inline comments on images for focused edits; and prompt sharing so others can remix the idea with their own photos.
For developers, OpenAI launched two API models: GPT-Image-2.5 Flare, the default general-purpose option offering higher quality than GPT-Image-2 at 50% lower latency; and GPT-Image-2.5 Sunburst, targeted at premium production-grade visual workflows with tighter control and longer generation times.