OpenAI has released ChatGPT Images 2.5, a new flagship image model for ChatGPT and its API, built around sharper detail, more obedient editing and speed. The company says generation latency drops by up to 50 percent compared with Images 2.0.
Lighting reads more naturally and textures come out richer; subjects from a reference photo stay recognizable when dropped into new scenes, and edits hold up across several rounds of instructions rather than fading under later changes. OpenAI describes the weekly volume of images created through ChatGPT and its image APIs as more than 3B.
New product features in ChatGPT aim at control. Sketch lets users draw directly in the chat window as a reference for the final image, templates speed up common formats such as flyers and product shots, and users can now drop comments directly on an image to target an edit. Prompts can be shared so others can reuse an idea with their own photos.
The update reaches the entire ChatGPT product family, plus Work and Codex tiers, across desktop, mobile and web. Developers get two API models: GPT-Image-2.5 Flare, which mirrors the consumer improvements, and GPT-Image-2.5 Sunburst, a slower, higher-precision option for detailed creative work where a single element, like a product or background, must change while the rest of the composition stays intact.
OpenAI stressed its safety work on the release, including measures around likeness and content restrictions.
