Google has unveiled Gemini 2.5 Flash Image (nicknamed “nano-banana”), its latest image generation and editing model, aimed at giving developers greater creative control and higher-quality outputs.
The update builds on Gemini 2.0 Flash, which was praised for its low latency and cost-effectiveness but had limitations in image quality and editing precision. Gemini 2.5 Flash Image addresses these gaps with several new features:
Multi-image fusion – Blend objects, restyle environments, and combine multiple photos into a single realistic image.
Character consistency – Maintain the same character or product appearance across different settings and angles, useful for storytelling, branding, and catalogs.
Targeted transformations – Make precise edits via natural language prompts, such as blurring backgrounds, removing objects, altering poses, or colorizing photos.
World knowledge integration – Use Gemini’s semantic understanding to read hand-drawn diagrams, assist with education, and follow complex editing instructions.
Template adherence – Generate uniform assets like employee badges, real estate cards, or dynamic product mockups.
Developer tools and pricing
Gemini 2.5 Flash Image costs $30 per 1 million output tokens, with each image priced at approximately $0.039. Google AI Studio has also upgraded its build mode, allowing developers to quickly test, remix, and deploy AI-powered apps—or save their projects to GitHub.
Partnerships and accessibility
To expand adoption, Google has partnered with OpenRouter.ai, bringing the model to its 3 million developers, and fal.ai, a platform for generative media. Gemini 2.5 Flash Image also incorporates SynthID, Google DeepMind’s invisible watermarking tool, to label all AI-generated or edited images.
Google said it is continuing work on improving long-form text rendering, factual detail accuracy, and further boosting reliability in character consistency. The model is expected to move from preview to stable release in the coming weeks.
