OpenAI Officially Launches GPT Image 1.5 a Comprehensive Upgrade to Its Image Editing Capabilities

OpenAI officially launched GPT Image 1.5 today, a new AI model optimized specifically for image generation and editing tasks. GPT Image 1.5 not only achieves significant improvements in generation speed and cost efficiency but also delivers more stable performance in text rendering, detail preservation, and multi-step image editing.

A New Model Designed Specifically for Image Generation

The release of GPT Image 1.5 comes at a time of increasingly fierce competition in the field of image generation models. Previously, Google had just launched its new generation image model, Nano Banana Pro, which showed advancements in text rendering and instruction execution. OpenAI's latest update clearly focuses on strengthening these key capabilities.

Compared to its predecessor, GPT Image 1.5 features significant improvements in the following areas:

• More precise text rendering capabilities
• Better execution of continuous, multi-step editing commands
• Preserves key details during image modification, reducing "over-editing" and "incorrect editing" issues

A Faster And More Resource-Efficient Image Generation Experience

OpenAI claims that switching models has improved the speed of ChatGPT's ability to generate images (4 times faster than before), which translates to a more efficient use of hardware resources (i.e., less powerful computers).

Reduced computing power = Faster response time = Reduced cost to operate the model (less time to perform task = less energy to operate). This will have a significant impact on development as well as on usage of platform-level applications.

The pricing of the API for GPT Image 1.5 is approximately 20% lower than OpenAI's previous image generation APIs (and developers can access the model via API calls, while people who do not know how to code will be able to access it via ChatGPT's built-in image generation capability).

Image Editing Capabilities More Suitable for Business Scenarios

In practical use, one of the common problems with image editing models is that modifying one element can inadvertently alter content the user didn't intend to change. While this has limited impact for ordinary users, it can pose serious risks in commercial scenarios such as branding, marketing, and e-commerce.

OpenAI points out that GPT Image 1.5 is better at preserving important visual elements during the editing process, such as brand logos and structural layouts. At the same time, its ability to understand complex instructions has also improved, for example:

Synthesizing objects from three different images into a single new image and uniformly adjusting the drawing style.

These types of multi-step, cross-image editing tasks are where GPT Image 1.5 significantly outperforms previous models.

Text Rendering Capabilities Have Been Significantly Improved

One of the greatest challenges faced by models in the past has been producing text as part of the generated images. OpenAI placed a heavy emphasis on improving its text-writing abilities during the training of the GPT Image Version 1.5. 

When making the announcement regarding the release of GPT Image, one of the major highlights was that GPT Image Version 1.5 produces images with a much greater degree of accuracy than past versions of GPT in that it now provides users with accurate text rendering capabilities for small, dense amounts of text. Examples of how this allows users to create infographics, data visualisation images and numerous types of visual content that have large amounts of text are shown. 

The release of GPT Image Version 1.5 enhances the ability of individuals to use the GPT Image product for professional purposes and commercial design.

Limitations of GPT Image 1.5

Certainly, GPT Image 1.5 isn’t flawless. In fact, OpenAI states in a recent blog that they know their model still has a very limited capability for art styles; plus, it can still produce incorrect images when it generates images from advanced scientific knowledge.

Nevertheless, these limitations are substantially better than previous versions, particularly with regard to improvements made relative to image stability on a broad range of science and technology areas.

Model Upgrade Schedule Proceeding In Parallel With GPT-5.2

It’s interesting to note that GPT image 1.5 came out just after OpenAI updated ChatGPT with GPT 5.2 as the base model.

According to OpenAI, the new GPT 5.2 has made significant strides in scientific reasoning tasks, achieving a record-breaking number of milestones in an internal testing environment that includes university-level mathematics. As part of its efforts to evaluate the performance of foundational models in scientific applications, OpenAI has also introduced a new AI benchmark called FrontierScience that is designed to help measure the performance of foundational models in scientific disciplines such as physics, chemistry, and biology.

This is an example of where OpenAI is currently focused with regards to their product offerings: In addition to increasing the capabilities of their foundational models, they are developing more focused and efficient models that meet specific vertical requirements (e.g., image generation).