OpenAI officially launched a new image generation model gpt-image-1.5 today, clearly targeting the much-anticipated Nano Banana series under the Google Gemini system to enhance its competitiveness in the field of AI image generation. This new model will be open to both developers and ordinary users, and will fully take over the image generation function in ChatGPT.
In March this year, OpenAI brought a major upgrade to the image generation function of ChatGPT. The new model quickly became popular once it was launched. Within a week, users generated more than 700 million images in ChatGPT. Then in April, OpenAI opened the same generation of image technology to developers by releasing the gpt-image-1 interface, enabling it to integrate image generation capabilities in various applications and services.
However, the Google Gemini Nano Banana model, which debuted in September, quickly seized the topic and market. With its stronger image generation and editing performance, it significantly increased Gemini's user coverage. Nano Banana is considered to be comprehensively ahead of gpt-image-1 in terms of picture quality and editing flexibility, and has been widely adopted by developers around the world. Last month, Google launched Gemini 3 Pro Image (codenamed Nano Banana Pro) on this basis, leveraging Gemini's advantages in reasoning and real-world knowledge to further improve the accuracy and consistency of generated images.
gpt-image-1.5, launched under this background, is regarded by OpenAI as a new generation core product that directly competes with the Gemini Nano Banana series. The new model has significantly improved the sophistication of image editing and the fidelity of key elements such as brand logos and faces. It is also more accurate in understanding and executing user instructions. OpenAI also emphasized that the new model has significantly enhanced the processing of text in images, especially the clarity and correctness of dense text and small font size text.
For developers, cost optimization is a highlight of gpt-image-1.5: compared with the previous generation gpt-image-1, the new model is about 20% cheaper in billing for image input and output. Developers can still finely control the generation cost through the quality parameter, and OpenAI claims that the new model can still maintain a strong generation effect even at lower quality settings.
For ordinary users, gpt-image-1.5 has become the underlying engine for ChatGPT's built-in image generation experience. OpenAI said that this model can perform precise editing while keeping details such as lighting, composition, and character characteristics basically unchanged. Compared with the previous generation, it can achieve up to 4 times performance improvement, and the generation and editing response speed is faster.
In conjunction with the new model, OpenAI has also launched a new "Images" image area in ChatGPT, which has been opened simultaneously in the mobile application and web version. Users can use preset filters and prompt words to quickly get started with image generation in this area. These presets are regularly updated to follow the latest creative styles and trends. OpenAI stated that the new image generation model and ChatGPT Images experience will be gradually pushed to all ChatGPT users, and it will continue to iterate around ease of use and creation efficiency in the future.