OpenAI launches ChatGPT Images 2.5 with faster editing and new creation tools

OpenAI shipped ChatGPT Images 2.5 to all ChatGPT, ChatGPT Work, and Codex users, focusing on speed and edit fidelity. The update claims up to 50% lower generation latency, sharper details with more natural lighting and richer textures, better preservation of reference photos, and more reliable multi-turn editing — the recurring pain point where iterative edits previously drifted away from the original subject.

On the creation side, the headline addition is 'Sketch,' a tool that lets users draw directly inside ChatGPT to guide generation, alongside new templates. Greg Brockman framed it as 'another step for creating whatever you can dream of.' The practical thrust is workflow: faster round-trips and steadier edits matter more to real users than raw novelty, especially for design and marketing tasks that involve many small corrections.

Competitively, image generation is now table stakes across the majors, and the differentiator has shifted from 'can it make an image' to editing precision, latency, and integration into work surfaces (Codex, ChatGPT Work). The caveat is that image tooling has become a vector for abuse and IP disputes, and faster, more realistic edits raise the same authenticity questions NVIDIA is trying to address with synthetic-video detectors. Watch how Sketch adoption compares to competing canvas-style tools and whether the latency gains hold under the same capacity pressure squeezing Astra.