OpenAI launches ChatGPT Images 2.5 with Sketch

OpenAI is positioning this update as an enhancement for iterative creative work. The previous version, ChatGPT Images 2.0 launched in April, had a weakness: when users made targeted edits in complex scenes with multiple subjects, the details would often degrade. The new 2.5 model reportedly addresses this, maintaining overall image stability so that users can make pinpoint edits.

The introduction of two distinct models—GPT-Image-2.5 Flare and GPT-Image-2.5 Sunburst—reflects a tradeoff between speed and precision. Flare generates images two to four times faster and supports transparent backgrounds, which is useful for brand assets and presentations. Sunburst is slower but offers greater fidelity, which the company says is important for creative professionals and artists who need pixel-perfect consistency in product photos and marketing materials.

The security measures are likely to be a key point for businesses considering adoption. OpenAI says ChatGPT Images 2.5 has guardrails on prompts and outputs, and uses C2PA metadata and invisible watermarking to identify generated images. This could matter for companies concerned about deepfakes and unauthorized reuse. The availability of inline comments and prompt sharing also suggests OpenAI is targeting collaborative workflows where teams iterate on images together.