Citation and evidence

Nano Banana Pro

12 min full readUpdated 15 references

This article's verification

Report a problem with this article

Edit article
More

Use this article

Raw MarkdownExplore connections

Improve this page

Revision historyDiscuss this article

Topics

Generative AIImage GenerationMultimodal AI

Cite this article

Nano Banana Pro is a professional grade image generation and editing model from Google DeepMind, released on November 20, 2025. Nano Banana Pro is the consumer facing nickname for a model that Google officially calls Gemini 3 Pro Image. It builds on the Gemini 3 Pro reasoning model, and it is the higher end sibling of the original Nano Banana, the viral image tool that Google launched a few months earlier as Gemini 2.5 Flash Image. The Pro version keeps the playful name but aims at a more demanding audience, with headline features that include output up to 4K resolution, much sharper text inside images, the ability to ground a picture in live data from Google Search, and consistency across many reference images. [1][2][3]

Google DeepMind logo
Google DeepMind logo. Source: Wikimedia Commons, Public domain.

The original Nano Banana spread quickly after its August 2025 debut. It drove a wave of viral edits, from figurine portraits to retro photo styles, and Google said in October 2025 that it had already generated more than 5 billion images. [1][4][12] That popularity is part of why the Gemini app shot up the app store charts that autumn. [4] Nano Banana Pro is the answer to what those users asked for next, which was more resolution, more control, and text that actually reads correctly. It sits inside the broader family of text-to-image models and modern AI image generation systems, and it ships across Google's consumer, developer, and enterprise products at the same time.

How it relates to Nano Banana and Gemini 3 Pro

There are two Nano Banana models, and the difference matters. The base Nano Banana runs on Gemini 2.5 Flash, a fast and cheap model tuned for quick edits and casual creation. Nano Banana Pro runs on Gemini 3 Pro, the larger reasoning model from the Gemini line. [1][3] Because it inherits that reasoning ability, Nano Banana Pro can do a short planning step before it draws anything. Google describes this as a thinking process, where the model works through composition and content first and can produce up to two interim images on the way to the final image. [15]

At launch the split showed up in the Gemini app's model picker: the faster default used the base Nano Banana, while choosing the Thinking model routed the request to Nano Banana Pro. [6] On February 26, 2026 Google released Nano Banana 2 (Gemini 3.1 Flash Image), which replaced Nano Banana Pro across the Fast, Thinking and Pro models in the Gemini app, while Google AI Pro and Ultra subscribers kept access to Nano Banana Pro by regenerating an image from the three-dot menu. [11] As of November 2026 Google describes Nano Banana 2.1 as its most advanced image model in the Gemini app, with Nano Banana Pro still offered through a "Redo with Pro" option. [10] So the base model is not a stale duplicate that the Pro replaced. The two coexist, with the base aimed at speed and volume and the Pro aimed at quality and accuracy. People who want a fast meme edit can stay on the base, and people who want a legible 4K poster reach for the Pro.

"Meet Nano Banana Pro: Next-Level AI Image Generation & Editing", official video from Google, published November 20, 2025. Watch on YouTube.

Resolution and output

Nano Banana Pro generates images at 1K, 2K, and 4K resolution, which is a clear step up from the single lower resolution tier of the base model. [2][10] It supports a wide set of aspect ratios, from square 1:1 through portrait 9:16 and widescreen and cinematic formats, so the same model can produce a square social post, a tall phone wallpaper, or a wide banner. [2][5]

The higher resolution is the feature people notice first, because it makes the output usable for print mockups, presentation slides, and detailed scenes where small elements would smear at lower resolution. The 4K tier costs more to run, which Google reflects in both token counts and API pricing, covered in the availability section below.

Text rendering inside images

Legible text has been a long standing weakness for image models, and it is one of the areas where Nano Banana Pro improved the most. The model can render longer passages of readable text directly in a picture, rather than the garbled lettering that older systems produced. [2][3] It handles multiple languages, which helps with localization, and it can style text with specific fonts, textures, and calligraphy effects. [2]

This turns the model into a practical tool for posters, greeting cards, product mockups, diagrams, and menus, where the words have to be correct and not just decorative. Independent reviewers singled out text rendering as one of the clearest improvements over the original Nano Banana. [8][9]

Search grounded generation

The most distinctive capability is grounding an image in real information through Google Search. Nano Banana Pro can pull live, factual data and then lay it out as an infographic, a diagram, or a data visualization. [1][5] Because the model checks real sources, the numbers and labels in a chart are more likely to be accurate rather than plausible looking but wrong, which is the usual failure mode for AI generated infographics. [8]

Google frames this as connecting image creation to Gemini's world knowledge and reasoning, so a prompt can ask for something like a weather summary, a sports recap, or a recipe card and get a visual that reflects current real data. [1] Reviewers treated the more accurate infographics as the feature that sets the model apart from rivals. [8] The accuracy still depends on what Search returns, so it reduces errors rather than removing them entirely. [8]

Reference images, consistency, and studio controls

Nano Banana Pro can take up to 14 reference images as input and blend them into a single composition. [1][2][10] It can keep up to 5 people looking like themselves across different generations, which is useful for storyboards, brand characters, or any project that reuses the same faces. [1][5] Within that budget Google says the model can hold up to 6 objects at high fidelity while drawing on the wider set of inputs. [5] That is a meaningful jump in how much visual context the model can juggle at once.

On top of that it offers studio style editing controls. Users can change the camera angle, adjust focus and depth of field, apply color grading, and transform the lighting of a scene, for example turning day into night or adding bokeh. [1][2] Edits can be localized to one part of an image instead of regenerating the whole frame. [6] These controls move the model closer to a directable production tool than a single shot generator.

SynthID and provenance

Every image from Nano Banana Pro carries an invisible SynthID watermark, the digital signature Google DeepMind uses to mark AI generated content. [1][2] The Gemini app also lets people upload an image and ask whether it was made or edited by Google's AI, which checks for that SynthID signal. [1]

There is also a visible mark. Images made on the free and Pro tiers include a visible Gemini sparkle logo in the corner. Google AI Ultra subscribers, and developers using the API, can generate images without that visible logo, while the invisible SynthID watermark stays in place regardless. [6] In other words the visible badge is a tier feature, but the provenance signal is always present.

Availability and pricing

Nano Banana Pro launched across many surfaces at once. For consumers it is in the Gemini app, in NotebookLM, and in Workspace apps including Google Slides and Google Vids, and it is built into Google Ads. [1] In Search, it launched in AI Mode in the U.S. for Google AI Pro and Ultra subscribers. [1] For developers it is available in the Gemini API and in Google AI Studio, and for enterprise it launched on Vertex AI, with Gemini Enterprise support announced as coming soon, as well as Google's agentic development platform Antigravity. [1][5]

At launch, access in the Gemini app followed the subscription tiers: free users received a limited quota before reverting to the original Nano Banana, and Google AI Plus, Pro, and Ultra subscribers received higher limits. [1][4][6] Since Nano Banana 2 replaced it as the Gemini app default in February 2026, Nano Banana Pro is used there by regenerating an image with it, an option Google lists for paid Google AI subscribers. [10][11] Search's AI Mode integration is gated to the paid Pro and Ultra tiers. [6] At launch the API model was named gemini-3-pro-image-preview, and developers reached it through the Gemini API and Google AI Studio. [5][9] Google released the stable model ID gemini-3-pro-image on May 28, 2026 and shut down the preview ID on June 25, 2026. [13] Pricing runs on tokens rather than a flat per image rate. Image output is billed at 120 dollars per 1 million tokens. [14] A 1K or 2K image uses about 1,120 tokens and costs roughly 0.134 dollars, while a 4K image uses about 2,000 tokens and costs roughly 0.24 dollars. [8][9] That makes the highest resolution close to double the price of the standard one, so cost scales with how much detail a project needs. For comparison, a base Nano Banana image runs about 0.039 dollars, so the Pro model is several times more expensive per picture. [8]

Specifications and capabilities

AttributeDetail
DeveloperGoogle DeepMind
Official model nameGemini 3 Pro Image
Consumer nameNano Banana Pro
Base modelGemini 3 Pro
Release dateNovember 20, 2025
API model IDgemini-3-pro-image (stable since May 28, 2026); launched as gemini-3-pro-image-preview
Resolutions1K, 2K, 4K
Aspect ratiosMultiple, from 1:1 square to 16:9, 9:16, and ultrawide
Reference imagesUp to 14
Person consistencyUp to 5 people
Text renderingLong, legible, multilingual passages
Search groundingYes, via Google Search for factual infographics
Studio controlsCamera angle, focus, depth of field, color grading, relighting
ProvenanceInvisible SynthID watermark, plus visible logo on free and Pro tiers
API pricing0.134 dollars per 1K/2K image, 0.24 dollars per 4K image

Expanded article table

Reception and competition

Early coverage treated Nano Banana Pro as a serious upgrade rather than an incremental one. TechCrunch reported Google's claims that the model improves on its predecessor with more detailed images and accurate text in different styles, fonts, and languages, and that it is slower and costlier than the original. [7] Independent reviewer Simon Willison called it the best available image generation model, and developer Max Woolf wrote that it is the best AI image generator, with caveats, singling out the text quality and the higher resolution while flagging cost and the visible watermark. [8][9]

The model lands in a crowded field. It competes most directly with OpenAI's GPT Image, the system behind image creation in ChatGPT, and with ByteDance's Seedream 4.0, which also pushed into high resolution multi reference generation in 2025. In December 2025, Max Woolf noted that OpenAI's newly released gpt-image-1.5 was ahead of Nano Banana Pro on LMArena's text-to-image leaderboard but had difficulty with prompt adherence. [8] Against those rivals, Google's pitch leans on the Search grounding and the Gemini reasoning underneath, which competitors do not offer in the same way. The Imagen 4 line and earlier systems like DALL-E sit in the same broad market but target different tradeoffs of speed and cost.

Limitations

Nano Banana Pro is powerful but not flawless. Search grounding lowers the rate of wrong facts in infographics, yet the output is only as reliable as what Search returns, so charts can still contain inaccuracies and should be checked before anyone trusts them. [8] The higher usage limits sit behind paid plans, and the free tier caps how many images a person can make. [1][6] The visible Gemini logo on free and Pro images is fixed unless a user moves to the Ultra tier or the API. [6]

The model also carries the usual caveats of AI image generation. It can still misread an unusual prompt, and the thinking step adds latency compared with the faster base Nano Banana, so it trades some speed for quality. The API model launched as a preview, and Google replaced it with the stable gemini-3-pro-image model in May 2026; as of November 2026 the published price is still 0.134 dollars per 1K/2K image and 0.24 dollars per 4K image. [13][14]

Continue reading

More to read

Topics:Generative AIImage GenerationMultimodal AI

References

  1. ^1 ^2 ^3 ^4 ^5 ^6 ^7 ^8 ^9 ^10 ^11 ^12 ^13 ^14 ^15Google. "Nano Banana Pro: Gemini 3 Pro Image model from Google DeepMind." The Keyword, blog.google. November 20, 2025. blog.google/...nano-banana-pro
  2. ^1 ^2 ^3 ^4 ^5 ^6 ^7 ^8Google DeepMind. "Introducing Nano Banana Pro." deepmind.google. November 20, 2025. deepmind.google/...pro
  3. ^1 ^2 ^3Google Cloud. "Nano Banana Pro available for enterprise." Google Cloud Blog. November 20, 2025. cloud.google.com/...a-pro-available-for-enterprise
  4. ^1 ^2 ^3Jordan Novet. "Google launches Nano Banana Pro, an updated AI image generator powered by Gemini 3." CNBC. November 20, 2025. cnbc.com/...google-nano-banana-pro-gemini-3
  5. ^1 ^2 ^3 ^4 ^5 ^6Google. "Developers can build with Nano Banana Pro (Gemini 3 Pro Image)." The Keyword, blog.google. November 20, 2025. blog.google/...gemini-3-pro-image-developers
  6. ^1 ^2 ^3 ^4 ^5 ^6 ^7Abner Li. "Google rolling out Gemini 3-powered Nano Banana Pro image gen, editing." 9to5Google. November 20, 2025. 9to5google.com/...gemini-3-nano-banana-pro
  7. ^TechCrunch. "Google releases Nano Banana Pro, its latest image-generation model." TechCrunch. November 20, 2025. techcrunch.com/...ts-latest-image-generation-model
  8. ^1 ^2 ^3 ^4 ^5 ^6 ^7 ^8 ^9Max Woolf. "Nano Banana Pro is the best AI image generator, with caveats." minimaxir.com. December 2025. minimaxir.com/...nano-banana-pro
  9. ^1 ^2 ^3 ^4Simon Willison. "Nano Banana Pro aka gemini-3-pro-image-preview is the best available image generation model." simonwillison.net. November 20, 2025. simonwillison.net/...nano-banana-pro
  10. ^1 ^2 ^3 ^4Google. "Nano Banana 2 - Gemini AI image generator & photo editor" (Gemini app image generation overview, accessed November 2026). gemini.google. gemini.google/...image-generation
  11. ^1 ^2Google. "Nano Banana 2: Combining Pro capabilities with lightning-fast speed." The Keyword, blog.google. February 26, 2026. blog.google/...nano-banana-2
  12. ^Hillary K. Grigonis. "Nano Banana has generated 5 billion images in two months. Now, it's about to come to Google Photos." Digital Camera World. October 15, 2025. digitalcameraworld.com/...to-come-to-google-photos
  13. ^1 ^2Google AI for Developers. "Gemini deprecations." Gemini API documentation. Accessed November 2026. ai.google.dev/...deprecations
  14. ^1 ^2Google AI for Developers. "Gemini Developer API pricing." Gemini API documentation. Accessed November 2026. ai.google.dev/...pricing
  15. ^Google AI for Developers. "Nano Banana image generation." Gemini API documentation. Accessed November 2026. ai.google.dev/...image-generation

Improve this article

Add missing citations, update stale details, or write a clearer explanation. Anyone can edit: changes go live when you publish, and every edit is checked by an editor.

7 revisions · v7 · 2,424 words · full history

Fact-checks are independent of edits: a reviewer re-verifies the article against its sources and stamps the date. How we verify

Research and drafting on this wiki are AI-assisted, under named human editorial standards. How AI is used here

Reviewer note: Independent fact-check of the full article against its cited sources (62 claims, 17 sources fetched); 5 material + 11 minor fixed

Cite this page: AI Wiki. "Nano Banana Pro." aiwiki.ai, updated 28 Oct 2026, fact-checked 5 Nov 2026. CC BY 4.0. https://aiwiki.ai/wiki/nano_banana_pro

Edit article