OpenAI gave ChatGPT a major image-generation upgrade in early 2025 with a new model that quickly went viral. The model proved to be a major success for the company and reportedly helped bring millions of new users to ChatGPT.
Later, in April 2025, OpenAI brought the same image-generation technology to developers through the gpt-image-1 API. The company then followed it up in December 2025 with the release of the gpt-image-1.5, which introduced significant improvements over the previous version.
On the other hand, Google has been rolling out its Gemini Nano Banana series of image-generation models since September last year. Earlier this year, the company announced Nano Banana 2, also known as Gemini 3.1 Flash Image, a state-of-the-art image model that delivers Nano Banana Pro-level image quality with notable improvements.
To take on Gemini Nano Banana 2, OpenAI today introduced ChatGPT Images 2. During a livestream, OpenAI CEO Sam Altman and others showcased the new model’s capabilities. OpenAI stated ChatGPT Images 2 is significantly better at generating images that include text. For example, users can now create images of a macOS desktop window or a chat interface with text rendered much more accurately throughout.
OpenAI said Images 2.0 can generate images by following instructions more closely, preserving requested details, and accurately rendering fine-grained elements such as small text, iconography, UI components, dense compositions, and subtle stylistic constraints. The new model can also create images at up to 2K resolution in a range of aspect ratios, from as wide as 3:1 to as tall as 1:3.
There will be two versions of the Images 2.0 model:
- ChatGPT Images 2.0 instant
- ChatGPT Images 2.0 thinking
When a thinking or Pro model is selected in ChatGPT, Images 2.0 can refer to the web for real-time information related to a query and then generate more accurate images. It can also create multiple distinct images from a single prompt and double-check its own outputs.
Finally, Images 2.0 has stronger multilingual understanding and is now much better at rendering non-Latin text, including Japanese, Korean, Chinese, Hindi, and Bengali.
The gpt-image-2 model is available through the API for developers with the following pricing:
- $8.00 for input
- $2.00 for cached input
- $30.00 for output
The ChatGPT Images 2.0 instant model is now available to all ChatGPT and Codex users, while the ChatGPT Images 2.0 thinking model is reserved for ChatGPT Plus, Pro, and Business users.
0 Comments
Load the comments and join the conversation!
Read the comments, ask the editors questions, show respect and join the conversation.