OpenAI has begun rolling out a new version of ChatGPT Images, powered by its latest image generation model, GPT Image 1.5, bringing faster image generation, more reliable instruction following and improved photo editing capabilities.

The update, which is available from Tuesday (Dec 16), is rolling out globally to all ChatGPT users and developers via the application programming interface (API). Business and Enterprise access to the new ChatGPT Images experience will follow at a later date, the company said.

Faster image generation and more precise edits

According to OpenAI, GPT Image 1.5 can generate images up to four times faster than its previous model, while making more precise edits that preserve key details such as lighting, composition and facial likeness.

When users request changes to an uploaded image, the model is designed to alter only the specified elements, keeping other aspects consistent across multiple edits. This allows for more practical photo touch-ups, clothing and hairstyle try-ons, as well as stylistic transformations that retain the essence of the original image.

OpenAI said the model supports a wide range of editing actions, including adding, removing, blending and transposing elements within an image, without degrading overall visual quality.

Stronger instruction following and text rendering

GPT Image 1.5 also improves how ChatGPT Images follows complex instructions, particularly for structured or multi-step image requests. This includes generating images with specific layouts, grids or relationships between objects.

Another key upgrade is enhanced text rendering. The model is better able to handle dense and small text, enabling use cases such as infographics, posters and mock newspaper layouts, where accuracy and legibility are critical.

“These improvements make the model more reliable for both creative exploration and professional design workflows,” OpenAI said in their blog post.

A new Images experience in ChatGPT

Alongside the new model, OpenAI is introducing a dedicated Images space within ChatGPT, accessible via the sidebar on mobile apps and chatgpt.com. The feature includes preset styles, filters and trending prompts designed to help users quickly explore ideas without needing to write detailed prompts.

Users can also upload a likeness once and reuse it across future image creations, reducing the need to repeatedly select photos from their camera roll.

OpenAI said images now render in parallel, allowing users to continue generating new images while others are still processing.

Improvements, limitations and API access

OpenAI noted that while GPT Image 1.5 shows clear improvements over earlier versions, results remain imperfect in some areas. The company cited limitations in rendering multiple faces, certain artistic styles and multilingual text, though overall output quality and realism have improved.

In the API, GPT Image 1.5 offers the same upgrades as ChatGPT Images and is positioned for marketing, e-commerce and brand design use cases. Image inputs and outputs are priced about 20 per cent lower than the previous GPT Image 1 model, OpenAI said.

Companies including Wix, Canva, Envato and Figma Weave are already using the model in production, according to OpenAI.