[AIbase Report] With the release of macOS Tahoe26, iPadOS26, and iOS26, Apple's image generation app Image Playground has received a major update. Not only has it integrated ChatGPT as an image provider, but it also lays the foundation for introducing more third-party models in the future.

Initially, Image Playground used only Apple's own models, allowing users to create illustrations through preset themes such as "birthday" and "fantasy." However, due to its relatively basic generation results, Apple expanded its collaboration with OpenAI. In the new system, users can directly use ChatGPT's preset filters, or choose to call OpenAI's model directly, to achieve a richer variety of image styles and higher quality generation results.

QQ20250923-112918.png

More third-party models are coming soon

According to an analysis of the latest beta code by 9to5Mac, Apple is expanding support for third-party models beyond OpenAI in Image Playground. The new "estimated latency" metric and "provider identifier" added in the beta version suggest that Apple may select the best model based on real-time latency for each task, or use it for internal data analysis. However, more importantly, this clearly indicates a change in the app's underlying architecture, which will support multiple different AI image generation models.

QQ20250923-112925.png

Currently, although it is unclear which models will be among the first to join, a strong candidate is Google's Gemini2.5Flash Image, also known as "Nano Banana." This model has recently gained popularity for its surrealistic style and ability to generate realistic human portraits based on user images, and has helped Google's Gemini app rank highly in app store charts across many countries.