The name that stuck
"Nano Banana" started as an internet nickname for an anonymous, unusually good image model that showed up on a public model leaderboard in the summer of 2025. When Google confirmed it was theirs, the name was already viral β so Google kept it. Officially, Nano Banana is Gemini 2.5 Flash Image, Google's fast image generation and editing model in the Gemini "Flash" family.
The launch was not subtle. Between August 26 and September 9, 2025, Google's Gemini app added more than 23 million new users and people generated over 500 million images in two weeks. By October 2, 2025, Google had moved the model to general availability with broader aspect ratio support.
What made it stick wasn't raw image quality alone β plenty of models produce pretty pictures. It was that Nano Banana treated image generation as a conversation. You could upload a photo, ask for one specific change in plain English, and get back the same image with exactly that change β no masks, no layers, no re-rolling the whole composition to fix one detail. It could also blend multiple images together and keep a character, product or mascot looking identical across edits, which is the part professional users actually cared about.
Since that first release, the Nano Banana name has grown into a small family of models. Here's what each one actually is.
Nano Banana Pro β the flagship
Announced November 20, 2025, Nano Banana Pro is Google's nickname for Gemini 3 Pro Image, built on the Gemini 3 Pro reasoning model rather than the faster Flash line. It's a meaningfully different tier of model, aimed at work where the image is the deliverable:
- Industry-leading text rendering β legible, correctly-spelled text inside the image itself, including long passages and multilingual layouts, which is historically one of the hardest things for diffusion-style image models to get right.
- Identity preservation across up to 5 people in a single image, and the ability to blend up to 14 reference images into one generation.
- Studio-grade creative controls β localized edits, lighting and focus adjustments, camera angle changes and color grading, on top of native 1K / 2K / 4K output.
- Grounding with Google Search, so prompts that depend on real-world facts (a current weather chart, a specific building, a piece of factual text) can pull in accurate information instead of hallucinating it.
On this platform, Nano Banana Pro is the default model behind the Nano Banana Pro tool and the Upscale tool. Because it's the higher-capability model, it also costs more per generation β 12 credits, versus 4 for the standard tier (more on that in our credits guide).
Nano Banana 2 β the new default
On February 26, 2026, Google shipped Nano Banana 2, its nickname for Gemini 3.1 Flash Image. It isn't a replacement for Nano Banana Pro β it's the successor to the original Nano Banana, sitting in the fast, free-tier-friendly Flash lineup. The pitch is straightforward: image quality that gets close to Nano Banana Pro, at roughly twice the speed and about half the cost. It generates anywhere from 512px thumbnails up to full 4K output, across a long list of supported aspect ratios, and inherited the improved text rendering and natural-language editing that made the original model popular.
Within days of launch it became the default image model across the free and paid tiers of the Gemini app, plus Google Search's AI mode, Lens, AI Studio and Vertex AI β which tells you how confident Google is that most day-to-day image requests don't need the Pro tier at all.
On this platform, Nano Banana 2 is the default model behind the AI Image Generator, Image to Image, Inpainting, Remove Background and Style Transfer tools. At 4 credits per generation, it's the model you'll use for the large majority of everyday image work.
So which one should you use?
A simple rule of thumb:
- Default to Nano Banana 2 for product shots, social content, quick edits, background removal, style exploration and anything where you'll generate several variations before picking a winner. It's fast and inexpensive enough to iterate freely.
- Reach for Nano Banana Pro when the image is the hero asset β packaging, key visual, a poster or anything with text baked into the design, print-ready campaign art, or a shot where you need multiple people's faces to stay consistent and correct.
Both models sit behind the same simple interface: describe what you want, or upload an image and describe the change. There's no separate "Pro workflow" to learn β just a model picker and a credit cost that scales with capability.
The Nano Banana family isn't the whole picture
It's worth saying plainly: Nano Banana, Nano Banana Pro and Nano Banana 2 are all image models. If you're looking to animate a still photo, generate video from a text prompt, or work with models like Kling, Veo or Sora, that's a separate lineup entirely β covered in our guide to choosing an AI video model.