A new AI image editing model called “nano banana” has been released by Google DeepMind, topping the LMArena leaderboard with its impressive abilities. The model is now available in the Gemini app, providing users with a more consistent and predictable experience.
The nano banana model builds upon Google’s earlier introduction of AI-powered image editing in Gemini. While previous models were competent, they lacked consistency across edits. The new model addresses this issue by remembering details instead of relying on unpredictable changes.
With the nano banana model, users can make multiple edits to an image without worrying about the results looking altered from the original source material. This opens up several interesting use cases, such as reimagining someone’s style or attire, changing their appearance into a different genre, like a matador or ’90s sitcom character.
The Gemini app now offers improved AI image editing capabilities, making it easier for users to achieve consistent and realistic results without the need for manual adjustments.
Source: https://arstechnica.com/ai/2025/08/google-improves-gemini-ai-image-editing-with-nano-banana-model