What is Nano Banana?

Nano Banana is Google's AI image generation model - the name the company eventually adopted, publicly, for the model first released as an unbranded preview inside Gemini. It produces photographic, illustrative, and graphic-design-quality images from natural-language prompts, and it does so fast enough to feel conversational: ask for an image, watch it appear, ask for a tweak, watch it update.

Architecturally, Nano Banana sits inside the Gemini family. It is not a separate product so much as a specialised model that Gemini routes to when you ask for an image. The same prompt surface that handles text reasoning, code, and tool use also handles image generation - which is why the experience inside Gemini is "type what you want, get what you asked for", without a dedicated mode switch.

Nano Banana is the most popular image generation model in the world.- The Keyword (Google Blog)

The name, the codename, the brand

The story behind "Nano Banana" is part of why the brand stuck. Google launched the model quietly inside Gemini without naming it. The community on the LMArena leaderboard noticed an unidentified image model dominating its category and assigned it a placeholder nickname: nano-banana. The nickname took off.

Google could have given the model a corporate name - Imagen 4, Gemini Image, Pixel Studio - and asked the community to start using it instead. Instead, the company embraced the joke. "Nano Banana" became the official product name. Google's launch posts, blog announcements, and developer docs use it directly. The original codename is now the brand.

The naming choice tells you something useful about how Google is positioning the model: as a developer-and-creator product first, an enterprise marketing line second. Image generation is competitive and informal, and a name that lives on a leaderboard does better than one that lives on a slide deck.

What it does, in one paragraph

Nano Banana is general-purpose for image work. It generates from a natural-language prompt, edits the resulting image through follow-up turns rather than fresh generations, accepts reference images for style and subject consistency, renders legible in-image text (one of its standout strengths), and resizes a creation into different aspect ratios without cropping the parts you care about. What it does not do yet: generate video, generate audio, run as a local model. Those jobs belong to other Google models (Veo for video, Lyria for audio).

Nano Banana vs Nano Banana Pro

Google released two tiers in the same model family - now versioned as Nano Banana 2 on the standard tier and Nano Banana Pro on the professional tier (the model Adobe and other partners surface as "Gemini 3 with Nano Banana Pro"). The split is now familiar from how Anthropic ships Claude (Haiku / Sonnet / Opus) - same model, different sizes optimised for different trade-offs.

STANDARD TIER

Nano Banana (2)

  • The default image model inside Gemini, available across free and paid plans
  • Three speeds in the model menu - Fast, Thinking, and Pro
  • Fast, conversational, optimised for everyday creative work
  • Best for: ideation, social graphics, drafts, casual creation
PROFESSIONAL TIER

Nano Banana Pro

  • Available in Google AI Pro / Plus / Ultra, the paid Gemini API, and partner surfaces
  • Multi-reference consistency, fact-grounded infographics, higher resolution
  • Crystal-clear in-image text built for product mockups and posters
  • Best for: professional-grade, functional design - print, packaging, client work

For developers using the Gemini API directly, both tiers are accessible by selecting the right model identifier on the request. For consumer users, Gemini routes to the right tier based on subscription state, or you can hit "Redo with Pro" on any result to bump it up.

Where to use it

Nano Banana is reachable through several surfaces - both first-party and partner:

  • Gemini (gemini.google.com). The chat surface. Type a prompt, get an image. Continue the conversation to iterate. The way most people meet the model.
  • Google AI Studio. The browser-based developer playground. Useful for experimenting with prompts, comparing tier outputs, and building reference flows.
  • Gemini API. Programmatic access via the standard Gemini SDKs (JavaScript, Python, Go, more). The path for embedding image generation into your own product.
  • Vertex AI. Google Cloud's enterprise path - the same model, with the compliance and identity story Cloud customers need.
  • Adobe Firefly and Photoshop. Both tiers ship inside Firefly's Generate Image, Prompt to Edit, and Firefly Boards modules, and Nano Banana Pro appears in Photoshop's Generative Fill model picker as "Gemini 3 with Nano Banana Pro".
  • Workspace integrations. Gmail, Docs, and Slides surface Nano Banana-powered image features inline. The image generation lives where the work already happens.

The image-generation landscape

Nano Banana is one of a handful of competitive image models in late 2025 / early 2026. Each one has its own strengths and its own audience:

  • OpenAI GPT Image / DALL-E. Tightly integrated into ChatGPT and the OpenAI API. Strong general quality, particularly good at understanding complex prompts.
  • Midjourney. The artistic / aesthetic specialist. Discord-first interface, beloved by illustrators and creative professionals. Less interactive than the chat-native models.
  • Stable Diffusion ecosystem. Open-source models you can run yourself. The maximum-control option - LoRAs, fine-tunes, ControlNet, the whole technical toolbox.
  • Adobe Firefly. Tightly integrated into Adobe Creative Cloud. The default if your design pipeline already lives in Photoshop or Illustrator.
  • Nano Banana. Conversational, fast, exceptional at in-image text, the cleanest "talk to it" experience. The model most non-experts settle into.

None of these is universally best. The right choice depends on the workflow you already use, the kind of imagery you need, and whether you value control or speed at the moment of decision.

When to reach for it

Nano Banana is the right tool when:

  • You want to generate or edit an image inside a conversation you are already having with Gemini.
  • The image needs legible, well-rendered in-image text (posters, infographics, marketing graphics).
  • You are building image features into a Gemini API-based product and want one surface for text, vision, tool use, and images.
  • You already work in Google Workspace and want image generation to live where the rest of your content lives.

Reach for something else when you need a specific artistic style that another model is famous for (Midjourney), when you need on-device or offline generation (Stable Diffusion), or when your design pipeline is entirely inside Adobe Creative Cloud (Firefly).

For everything in the broad middle - the kind of "I need an image to go with this" work that comes up daily in marketing, product, and creative roles - Nano Banana is the model most people are reaching for now, and the one Google is investing the most in.