A mysterious AI image‑editing tool known as Nano Banana burst onto the scene in August 2025. According to AI‑Stack, the model first surfaced on the LMArena benchmark as an anonymous entrant and quickly impressed users with its natural‑language editing abilities, outperforming competitors and preserving character identities better than existing tools. User tests showed that the model could perform tasks such as adding or removing objects, changing backgrounds, completing faces and transferring styles, all with text prompts and without masking. This anonymous model achieved a 70 % win rate against other image‑editing systems and scored 0.89 on GenEval benchmarks, compared with DALL‑E 3’s 0.76.
On 26 August 2025, Google ended the mystery by confirming that the tool was indeed its creation. Axios reported that the company is integrating the model — officially named Gemini 2.5 Flash Image — into the Gemini app. Free users can perform up to 100 image edits per day, while paid users can perform ten times as many. Google highlighted that the model excels at multi‑step edits and image fusion; examples included transforming a portrait of a woman into a matador and applying a pattern from one image onto boots in another. The company says the model preserves facial consistency even after multiple edits, addressing a common limitation in AI image editors.
Â
