OpenAI has unveiled GPT Image 1.5, a new AI model specifically optimized for image generation tasks.
The release comes shortly after Google LLC launched its own advanced image generator, Nano Banana Pro, which boasts improved text rendering and better adherence to sequential prompts. GPT Image 1.5 matches these capabilities with further refinements and performance optimizations.
Developers can access GPT Image 1.5 via an API priced 20% lower than OpenAI’s previous image models. Non-technical users can interact with the model through ChatGPT's built-in image generation feature.
According to OpenAI, the integration of GPT Image 1.5 has accelerated image generation in ChatGPT by up to four times compared to earlier versions. This significant speed boost reflects enhanced hardware efficiency, reducing both computational demands and operational costs—faster processing means less energy consumption per task.
One common challenge in AI-driven image editing is the unintended alteration of elements not specified by the user. While minor in casual use, this can be problematic for commercial applications—brands resizing product images wouldn’t want logos automatically removed. OpenAI claims GPT Image 1.5 excels at preserving critical details during edits and handles complex, multi-step instructions more reliably. For instance, users can request combining objects from three different sketches into one image and then apply a unified stylistic transformation.
Text rendering remains a key focus area, mirroring one of Nano Banana Pro’s highlighted strengths. OpenAI reports that GPT Image 1.5 delivers superior performance in generating small, dense text blocks, making it particularly well-suited for creating detailed infographics and data visualizations.
The company acknowledges certain limitations in its official blog post. GPT Image 1.5 still struggles with some artistic styles and may produce inaccuracies when generating scientifically sensitive content. However, it shows marked improvement over prior iterations in these domains.
A new interface within ChatGPT provides access to the model, featuring preset filters and prompt suggestions that OpenAI plans to update regularly.
GPT Image 1.5 arrives just days after OpenAI upgraded ChatGPT to run on GPT-5.2, its latest foundational model. The updated core demonstrates stronger performance on scientific reasoning tasks, setting a new record in an internal benchmark involving university-level mathematics problems.
On the same day, OpenAI introduced FrontierScience, its proprietary AI benchmark designed to evaluate foundation models on scientific proficiency. The test suite includes over 700 text-based questions spanning physics, chemistry, and biology.
