What is Nano Banana? Google's AI image models explained
Nano Banana is the popular name for Google's Gemini image generation and editing models. The original Nano Banana is Gemini 2.5 Flash Image, released in August 2025; Nano Banana Pro is Gemini 3 Pro Image, Google's most advanced image model, released in November 2025; and Nano Banana 2 is Gemini 3.1 Flash Image, released in February 2026 to bring Pro-level features to Flash speed. All of them create and edit images from plain-language instructions and are known for keeping people recognisable from a single photo.
Where the name came from
Nano Banana started as a codename. Before launch, Google entered its new image model anonymously on LMArena, a site where people compare AI models side by side without knowing which is which. Asked for a codename in the middle of the night, a Google product manager combined two of her nicknames, and the model went up in early August 2025 as 'nano-banana'.
People started sharing results from the mysterious model, and the name stuck even though the technical name remained Gemini 2.5 Flash Image. Google has since used it for the whole family.
Nano Banana (Gemini 2.5 Flash Image)
Announced on 26 August 2025, the first Nano Banana made editing a photo by describing the change mainstream. Its headline abilities were blending several input images into one, keeping a character or object consistent across edits, making targeted changes from a sentence (blur the background, remove a person, change a pose, colourise a black-and-white photo) and using Gemini's knowledge of the world to follow multi-step instructions.
It became generally available in October 2025 and powered a wave of viral trends, from desk figurines to portraits with your younger self. For developers, Google now recommends its newer Nano Banana 2 Lite as the replacement, but the original is still fast and good at playful transformations.
Nano Banana Pro (Gemini 3 Pro Image)
Released on 20 November 2025 and built on Gemini 3 Pro, Nano Banana Pro is Google's most advanced image generation and editing model. It draws on Gemini 3 Pro's reasoning and world knowledge, which shows in the details:
- Output at 2K and 4K, in a range of aspect ratios.
- Legible text, from short taglines to long paragraphs, in several languages, including translating the text inside an image.
- Blending of up to 14 input images while keeping the resemblance of up to five people.
- Studio-style controls: change the camera angle, focus and depth of field, colour grading or lighting, for example turning day into night.
- Grounding in Google Search for real-world information such as the weather, useful for infographics and diagrams.
Nano Banana 2 (Gemini 3.1 Flash Image)
Nano Banana 2 arrived on 26 February 2026 with a simple promise: most of what Pro can do, at the speed of a Flash model. It outputs from 512 pixels up to 4K, renders and translates legible text, keeps up to five characters consistent and up to 14 objects in a single workflow, and can draw on Google web and image search. It replaced Pro as the default image model across the Gemini app and became the default in Flow, Google's video tool. Both Nano Banana 2 and Pro became generally available to developers in May 2026.
Two newer members followed. Nano Banana 2 Lite (Gemini 3.1 Flash Lite Image, June 2026) is Google's fastest and most cost-efficient image model, aimed at rapid drafts. Nano Banana 2.1, released in October 2026, is an update to Nano Banana 2 with better realism, text rendering and prompt adherence.
Which Nano Banana should you use?
Google itself recommends Pro when maximum factual accuracy matters and Nano Banana 2 for fast generation. In practice:
- Nano Banana Pro for the sharpest 4K results, posters and designs with text, and close portraits where every detail of a face matters.
- Nano Banana 2 for quick, high-quality edits and for iterating: try an idea at 1K, then render the keeper at 4K.
- The original Nano Banana for fast, playful transformations such as figurines and action figures.
Where you can use Nano Banana
Google offers the models in the Gemini app and Search, to developers through Google AI Studio and the Gemini API, and to businesses on Vertex AI. Third-party apps can use them through Google's API too. Luren AI is one of them: an independent app, not affiliated with Google, that offers Nano Banana, Nano Banana 2, Nano Banana 2 Lite and Nano Banana Pro alongside models from OpenAI, ByteDance and xAI, so you can choose the model for each image.
Prompting is the same wherever you use them: describe the scene in full sentences, say what must stay unchanged when you edit a photo, and refine conversationally ('now make it night'). The Nano Banana Pro prompts collection has tested examples, and the Nano Banana Pro vs GPT Image 2.5 comparison shows how Google's top model differs from OpenAI's.
Watermarks and AI labels
Every image created or edited by the Nano Banana models carries SynthID, an invisible digital watermark that identifies it as AI-generated, and you can upload an image to the Gemini app and ask whether Google AI made it. Google is also adding C2PA Content Credentials, an industry standard for recording how an image was made. When Pro launched, Google said images made in the Gemini app on free and Google AI Pro plans would also carry a visible sparkle mark.
