OpenAI ships ChatGPT Images 2.5 with Sketch and faster API models

8 Sep 2026: OpenAI published ChatGPT Images 2.5 — sharper detail, more precise multi-turn edits, and up to 50% lower generation latency vs Images 2.0 (OpenAI’s numbers).

8 Sep 2026: OpenAI published “Introducing ChatGPT Images 2.5.” That is the filing event. OpenAI: sharper details, more precise multi-turn editing, and “50% lower latency” versus GPT-Image-2 / Images 2.0 on the Flare API line. Those quality and latency figures are OpenAI’s.

OpenAI: available to all ChatGPT, ChatGPT Work, and Codex users on desktop, mobile, and web. The Verge’s same-day recap repeats that rollout.

New ChatGPT tools on the product page: Sketch — draw directly in ChatGPT as a reference for the final image (The Verge: type @Sketch to open the drawing window); templates (flyers, product photos); comments placed on images for focused edits; shareable prompts so others can retry an idea with their own photos and details.

API: GPT-Image-2.5 Flare (default / faster — “higher-quality images than GPT-Image-2 at 50% lower latency”) and GPT-Image-2.5 Sunburst (higher precision, longer generation times). Developer docs list gpt-image-2.5-flare and gpt-image-2.5-sunburst, with a dated Flare snapshot gpt-image-2.5-flare-2026-09-08.

OpenAI: “Every week, people create more than 3 billion images across ChatGPT Images and the GPT-Image models in the API.” That volume figure is the company’s.

Safety / provenance on OpenAI’s Images 2.5 system card (same day): C2PA metadata via the C2PA Conformance Program, plus invisible watermarking through Google DeepMind’s SynthID on ChatGPT, Codex, and the API. Those are OpenAI’s stated controls, not a third-party audit.

OpenAI put a dated product page on Images 2.5 with ChatGPT UX tools and two API variants. File the ship and OpenAI’s latency/volume claims as theirs.