LLMOcheki adopts GPT-Image-2.5, OpenAI's latest model, for image creation in article generation

Itera, Inc. (head office: Shinjuku-ku, Tokyo; Representative Director: Takayuki Muto) has moved image creation in the AI article generation feature of LLMOcheki, its tool for measuring and improving LLMO, onto GPT-Image-2.5 — the latest image generation model, released by OpenAI in the early hours of 9 September 2026.

The switch came within hours of the announcement. Faster generation and more precise editing shorten the time it takes to produce an article’s key visual and the diagrams inside it.

Background: the article gets written, then stalls on the images

AI article generation has reached a practical stage, and the time it takes to write the body has fallen sharply. Preparing the key visual and the diagrams in the body, however, has tended to stay manual — and that is where work piles up on the way to publication.

Search a stock site, reuse an existing image, or publish without images. Each of those sacrifices either the quality of the article or the speed of producing it.

What has changed

GPT-Image-2.5 support

According to OpenAI, the model shortens the wait for generation compared with the previous one, and is more precise at editing only the part you specify while leaving everything else intact. Handling of complex layouts, including transparent backgrounds, has also improved.

Already built into the production flow

LLMOcheki sets the key visual automatically and inserts images into the body of articles produced with the AI article generation feature. With this model update, those images are now generated in less time.

Provenance information for generated images

The image generation models OpenAI provides attach C2PA-standard metadata and an invisible watermark. Generated images remain identifiable as AI-produced.

On how we develop

LLMOcheki is set up to respond to updates in AI technology within a short time. Generative AI moves quickly and a model update bears directly on product quality, so keeping pace with external technical updates is treated as a development priority.

Availability

Available on plans that include the AI article generation feature, at 2 CR per generation.

Comment from our Representative Director

“People can write the body of an article with AI now, but we keep hearing that they stall on the images. When generation gets faster and you can fix only the part you want to fix, the path to publication changes. In this field a model update goes straight to the value of the product, so we keep ourselves set up to put one in as soon as it ships.”

LLMOcheki https://llmocheki.com/