Phaet
Gemini 3.1 Flash Image#1 Image Arena

Nano Banana 2

Pro quality at Flash speed. The #1 ranked model on Artificial Analysis Image Arena — 3-5x faster generation, up to 4K resolution, and advanced capabilities like image search grounding, all at roughly half the cost.

Powering image generation across Google Search, Gemini App, Flow & Ads

0/10
6 credits

Pro-Level Fidelity, Flash-Level Speed

Built on Gemini 3.1 Flash architecture, Nano Banana 2 achieves ~95% of Nano Banana Pro's image quality while generating 3-5x faster. It draws from Gemini's real-world knowledge base and live web search to render subjects with remarkable accuracy — from real landmarks to brand logos.

3-5x
Faster than Pro
~95%
Of Pro Quality
4K
Max Resolution
14
Reference Images
15+
Aspect Ratios
#1
Image Arena Rank

See What Nano Banana 2 Can Create

Real outputs generated with Nano Banana 2 — from photorealistic scenes to stylized illustrations

What Sets It Apart

Six core capabilities that make Nano Banana 2 the most versatile image model available

Image Search Grounding
EXCLUSIVE

Image Search Grounding

A capability exclusive to Nano Banana 2. The model searches Google Images for visual references during generation, dramatically improving accuracy for real-world subjects like landmarks, celebrities, and brand assets.

Precision Text Rendering

Precision Text Rendering

Currently the leading model for in-image text generation. Produces accurate, legible text with improved spacing and formatting — ideal for marketing mockups, posters, greeting cards, and infographics.

Multi-Character Consistency

Multi-Character Consistency

Maintain consistent appearance for up to 5 characters and fidelity across 14 objects in a single workflow. Perfect for comic strips, storyboards, brand mascots, and sequential narratives.

Style & Texture Transfer

Style & Texture Transfer

Extract texture, color palette, or artistic style from any reference photo and apply it to new subjects. Experiment with aesthetics like oil painting, watercolor, pixel art, or photorealistic rendering without starting from scratch.

Natural Language Editing

Natural Language Editing

Edit existing images with plain text instructions — swap backgrounds, remove objects, add elements, or change lighting conditions. The model preserves details you want to keep while precisely modifying what you describe.

Configurable Thinking

Configurable Thinking

Control the model's reasoning depth: Minimal for instant results, High for complex scenes requiring spatial reasoning, or Dynamic to let the model decide. Tune the speed-quality tradeoff for your specific workflow.

How It Compares

Nano Banana 2 delivers the best balance of speed, quality, and cost in the market

FeatureNano Banana 2Nano Banana ProGPT Image 1.5
Generation Speed 4-8 seconds 10-20 seconds 60+ seconds
Max Resolution 4K 2K 2K
Text Rendering Best in class Good Very good
Image Search Grounding Yes No No
Reference Images Up to 14 Up to 8 Limited
Aspect Ratios 15+ incl. ultra-wide Standard set Standard set
Configurable Thinking 3 modes Fixed Fixed

Built for Every Creative Workflow

From rapid iteration to polished final assets

Marketing & Ads

Marketing & Ads

Generate campaign visuals, social media assets, and A/B test variations in seconds. Text rendering makes banner ads with accurate copy a reality.

Comics & Storytelling

Comics & Storytelling

Create consistent characters across panels and scenes. Maintain the same protagonist through an entire visual narrative with multi-character consistency.

Product & Mockup Design

Product & Mockup Design

Visualize product concepts, packaging designs, and UI mockups with photorealistic quality. Use style transfer to explore different aesthetic directions rapidly.

Content & Social Media

Content & Social Media

Produce scroll-stopping visuals at the pace social platforms demand. Ultra-wide ratios are perfect for YouTube thumbnails, Twitter banners, and Pinterest pins.

Frequently Asked Questions

Start Creating with Nano Banana 2

The #1 ranked AI image model is ready. Pro-quality visuals in seconds, not minutes.