Google releases Gemini 2.5 Flash Image, nicknamed 'Nano Banana'

Google released Gemini 2.5 Flash Image, an image generation and editing model that used Google’s own nickname for it, “nano-banana,” in the announcement itself. The model could blend multiple source images into one, apply targeted edits described in natural language, and — its most distinctive claimed capability — maintain a consistent character or subject across a sequence of generated images, addressing a persistent weakness of earlier image models in which the same character would subtly change appearance between generations. Google said the model also drew on Gemini’s general world knowledge during generation and editing, rather than working from image data alone.

The model launched in preview on 26 August 2025 through the Gemini API, Google AI Studio for developers and Vertex AI for enterprise customers, priced at $30 per million output tokens, with each image counted as 1,290 output tokens — about $0.039 per image. Google said a stable release would follow within weeks. Third-party platforms including OpenRouter and fal.ai made the model available to their developer communities at launch.

The “nano-banana” nickname, which Google adopted rather than suppressed, spread quickly and became the model’s common name in coverage and developer discussion, ahead of the more conventional editing and generation model naming used elsewhere in the Gemini line. The release added Google to an active field of AI image-editing tools — including offerings from OpenAI, Midjourney and Black Forest Labs — competing specifically on consistency and multi-image composition rather than raw generation quality alone, a capability gap that had become one of the more visible differentiators between image models by late 2025.