Nano Banana is the name Google uses for its Gemini image generation and editing models, and the family now spans five model IDs with different prices, limits and retirement dates. Google's Gemini API documentation lists Nano Banana 2.1 (gemini-nano-banana-2.1) as the primary model, released as generally available on 6 October 2026, alongside Nano Banana 2 Lite, Nano Banana 2, Nano Banana Pro and the original Nano Banana (Gemini 2.5 Flash Image), which is now legacy. The models are available to developers through the Gemini API and Google AI Studio, to enterprises through Google Cloud, and to consumers through the Gemini app and Google Pics, a new image tool announced at Google I/O on 20 May 2026. This page sets out the model IDs, prices, features, provenance signals, data handling and dated release history from Google's own pages. It is written for developers, creative and marketing teams, and the security, privacy and procurement staff who approve AI tools.
Last checked: 2026-11-15.
This is an independent reference page from NHI Mgmt Group based on public provider documentation; it is not sponsored or endorsed by any provider. It draws on Google's public Gemini API documentation as of November 2026. Product features change, so confirm settings against Google's current documentation.
Key takeaways
- Five models, one brand. Nano Banana 2.1, Nano Banana 2 Lite, Nano Banana 2, Nano Banana Pro and the legacy Nano Banana (Gemini 2.5 Flash Image) each have their own model ID, price and limits.
- No free API tier for images. Google's pricing page lists none of the image models on the free tier; paid API output images cost from about USD 0.0336 (1K) to USD 0.24 (4K, Nano Banana Pro).
- SynthID on every image. Google states that all images generated through the Gemini API include an invisible SynthID watermark, and Google Cloud lists C2PA Content Credentials as supported for Nano Banana Pro.
- Retirements are scheduled. Preview IDs were shut down on 25 June 2026, Imagen 4 on 17 August 2026, and gemini-2.5-flash-image has an earliest shutdown date of 15 March 2027.
- Data use depends on the tier. Google's pricing page says free-tier content can be used to improve products and paid-tier content is not, while Gemini app chats follow the Gemini Apps activity settings.
What is Google Nano Banana?
Nano Banana is a family of natively multimodal Gemini models that generate and edit images from text, images and, on some models, video. Unlike Google's earlier Imagen models, which were dedicated text-to-image models and are now shut down, Nano Banana models are Gemini models that return images inside a normal Gemini API conversation, so they can take several reference images, follow multi-turn editing instructions and use Google Search results as context. Google describes Nano Banana 2.1 as "the primary high-efficiency workhorse model" and Nano Banana Pro as "the premium choice for the most complex visual tasks".
The name first appeared with Gemini 2.5 Flash Image, launched as a preview on 26 August 2025. Google's documentation now uses Nano Banana as the product name for every Gemini image model, with the technical model ID shown alongside.
Key facts
| Field | Value (as of 2026-11-15) |
|---|---|
| Provider | Google (Gemini API, Google AI Studio, Google Cloud, Gemini app, Google Pics) |
| Modality | Image output; text, image and (on some models) video input |
| Primary model | Nano Banana 2.1, gemini-nano-banana-2.1, GA 6 October 2026 |
| First release | Gemini 2.5 Image Preview, 26 August 2025 |
| Resolutions | 1K, 2K and 4K (Nano Banana 2.1 and Pro); 512px only on Nano Banana 2; 1K only on Nano Banana 2 Lite |
| Reference images | Up to 14 in total, split by model into object, character and style references |
| Open weights | No: API and app access only |
| Free API tier | Not available for image models |
| Watermark | SynthID on all generated images (Gemini API documentation) |
| C2PA Content Credentials | Listed as supported for gemini-3-pro-image on Google Cloud; Gemini app verifies C2PA credentials from 20 May 2026 |
| Knowledge cutoff | January 2025 for gemini-3-pro-image (Google Cloud model page); not stated for the other models on the pages checked |
Models and versions
The table lists the models in Google's Gemini API documentation, newest first. Google's deprecations page notes that shutdown dates are the earliest possible retirement dates and that exact dates are communicated in advance.
| Name | Model ID | Released (GA) | Status and shutdown | Google's description |
|---|---|---|---|---|
| Nano Banana 2.1 | gemini-nano-banana-2.1 | 6 October 2026 | Primary; no shutdown date announced | Update to Nano Banana 2 and primary workhorse model |
| Nano Banana 2 Lite | gemini-3.1-flash-lite-image | 30 June 2026 | Active; no shutdown date announced | Fastest and cheapest Gemini image model |
| Nano Banana 2 | gemini-3.1-flash-image | 28 May 2026 | Deprecated 6 October 2026, no shutdown date; replacement gemini-nano-banana-2.1 | Previous generation |
| Nano Banana Pro | gemini-3-pro-image | 28 May 2026 | Active; Google Cloud lists retirement 28 May 2027 or later | Premium choice for complex visual tasks |
| Nano Banana (legacy) | gemini-2.5-flash-image | 2 October 2025 | Legacy; earliest shutdown 15 March 2027; replacement gemini-3.1-flash-lite-image | Original Nano Banana |
As of 2026-11-15. Retired IDs in the same family include gemini-2.0-flash-preview-image-generation (shut down 14 November 2025), gemini-2.5-flash-image-preview (15 January 2026), and gemini-3-pro-image-preview and gemini-3.1-flash-image-preview (both 25 June 2026). Code that still references a preview ID fails, so treat the ID string as configuration that needs an owner. For the wider Gemini model line, see the Google Gemini Models.
Pricing and plans
API use is billed per million tokens, with output images converted to tokens. Google's pricing page also gives an approximate per-image price. All prices are in USD and none of the image models has a free tier.
| Model (as of 2026-11-15) | Input per 1M tokens | Text and thinking output per 1M | Image output per 1M | Approximate price per image | Batch |
|---|---|---|---|---|---|
| Nano Banana 2.1 | 1.50 (text, image, video) | 7.50 | 30.00 | 1K 0.0336; 2K 0.0504; 4K 0.113 | Image output 15.00; 1K 0.0168, 2K 0.0252, 4K 0.0567 |
| Nano Banana 2 Lite | 0.25 (text, image, video) | 1.50 | 30.00 | 1K 0.0336 | Image output 15.00; 1K 0.0168 |
| Nano Banana 2 | 0.50 (text, image) | 3.00 | 60.00 | 0.5K 0.045; 1K 0.067; 2K 0.101; 4K 0.151 | Not listed |
| Nano Banana Pro | 2.00 (text, image) | 12.00 | 120.00 | 1K and 2K 0.134; 4K 0.24 | Not listed |
| Nano Banana (legacy) | 0.30 (text, image) | Not listed per 1M | 30.00 | 0.039 up to 1024x1024 | Input 0.15; output 0.0195 per image |
Source: Google AI Studio pricing page, checked 2026-11-15. Two cost details are easy to miss:
- Thinking is billed: for Gemini 3 image models, thinking cannot be disabled in the API, and Google states that thinking tokens are billed by default. The interim "thought images" are not charged.
- Token conversion: on Google Cloud, gemini-3-pro-image counts 560 tokens per input image and 1,120 tokens per output image at 1K and 2K, rising to 2,000 tokens at 4K.
In the Gemini app, image generation draws on the plan's general usage limit rather than a fixed image count. Google's help page lists image generation with Nano Banana 2 on every tier, including no plan, and "Redo images with Nano Banana Pro" on Google AI Plus, Pro and Ultra. It sets limits as multiples of the standard limit: Plus 2x, Pro 4x, and Ultra 5x or 20x the Pro limit, and notes that media generation uses more of the allowance than simple chats.
Features
- Text-to-image and editing: generate from a prompt, or send an image with instructions to change it. Google recommends multi-turn chat for iterating on an image.
- Reference images: up to 14 inputs in total. Nano Banana 2.1 and Nano Banana 2 take up to 10 object and 4 character references; Nano Banana Pro takes up to 6 object, 5 character and 3 style references; Nano Banana 2 Lite takes up to 14 object references.
- Grounding: Google Search grounding on every current model except Nano Banana 2 Lite; Image Search grounding only on Nano Banana 2.1 and Nano Banana 2.
- Video to image: supported on Nano Banana 2.1, Nano Banana 2 and Nano Banana 2 Lite.
- Thinking levels: Nano Banana 2.1 supports minimal, medium and high (default medium); Nano Banana 2 and 2 Lite support minimal and high (default minimal).
- Interleaved output: Nano Banana Pro can return interleaved images and text, and Google Cloud lists virtual try-on and person generation among its supported features.
- Aspect ratios: Google Cloud lists 14 ratios for Nano Banana Pro, from 1:1 to 21:9 and including 1:8 and 8:1; the Gemini API page lists ten ratios for Nano Banana 2 Lite.
Two parameter rules cause avoidable errors: image sizes must use an uppercase K ("1k" is rejected), and 512px output is supported only on Nano Banana 2, not on Nano Banana 2.1.
Availability
| Route (as of 2026-11-15) | Who it is for | What Google states |
|---|---|---|
| Gemini API and Google AI Studio | Developers | All five models; paid tier only for images; Batch API with turnaround of up to 24 hours |
| Google AI Studio Build agent | App builders | Can generate custom images "on the fly" using Nano Banana (I/O 2026) |
| Google Cloud (Gemini Enterprise Agent Platform) | Enterprises | gemini-3-pro-image GA in the global location, with Provisioned Throughput, Standard, Flex and Priority pay-as-you-go |
| Gemini app | Consumers and subscribers | Nano Banana 2 on all tiers; Nano Banana Pro redo on Plus, Pro and Ultra |
| Google Pics | Consumers and Workspace customers | Built on the latest Nano Banana model; trusted testers from 20 May 2026, global rollout to Google AI Pro and Ultra subscribers announced for summer 2026, preview for Workspace business customers |
Google described Google Pics at I/O as an image creation and editing tool with object segmentation, text editing and translation, and Workspace integrations. Organisations that use Gemini through Workspace should read the Gemini for Google Workspace Security Guide for admin controls, since the consumer and work data terms differ.
Provenance and labelling
Google uses two separate provenance signals, and they answer different questions.
- SynthID watermark: the Gemini API documentation states that "All generated images include a SynthID watermark." Google DeepMind describes SynthID as an invisible watermark added when content is created and designed to withstand changes such as cropping, filters and lossy compression. It is checked by Google tools, not visible to the eye.
- C2PA Content Credentials: signed metadata recording how a file was made. Google Cloud lists Content Credentials (C2PA) as a supported feature of gemini-3-pro-image. The Gemini API image generation page does not state C2PA support for the other models.
- Verification: users can upload a file to the Gemini app, Search or Chrome and ask whether Google AI made or altered it. At I/O 2026 Google said SynthID verification in the Gemini app had been used 50 million times, was expanding to Search, and was coming to Chrome; C2PA verification began in the Gemini app on 20 May 2026, with Search and Chrome to follow. Google also runs a SynthID Detector portal.
Metadata can be stripped when an image is screenshotted, re-encoded or uploaded to some platforms, while a watermark is designed to survive common edits. Neither signal proves that an unmarked image is human-made. See SynthID vs C2PA Content Credentials for how the two compare, the AI Content Provenance and C2PA Guide for the standard itself, and the EU AI Act Article 50 Transparency Obligations Guide for labelling duties.
Data and privacy
| Route (as of 2026-11-15) | Used to improve Google products? | Retention stated |
|---|---|---|
| Gemini API free tier | Yes (pricing page); image models are not offered on it | Not stated on the pricing page |
| Gemini API paid tier | No (pricing page) | Not stated on the pricing page |
| Gemini app, Keep Activity on | Yes, including training generative AI models; some chats reviewed by humans | Default auto-delete 18 months (3 or 36 months selectable); reviewed data up to 3 years, disconnected from the account |
| Gemini app, Keep Activity off or temporary chat | No, unless feedback is submitted | 72 hours with the account |
| Work or school accounts | Different data handling terms may apply | See the Generative AI in Google Workspace Privacy Hub |
For the Gemini app, Google's privacy hub says uploaded images are processed with Google Lens technology and treated like any other prompt, and that gallery photos shared with Keep Activity on may be used to improve services with human reviewer help. Staff who upload product shots, customer photos or images of colleagues should know which route they are on. Uploading a photo of a real person for editing also raises consent questions; see the AI Image and Video Generation at Work guide for policy wording.
Where identity and security fit
Every API route depends on a credential. Gemini API keys created in AI Studio and Google Cloud service accounts that call gemini-3-pro-image are non-human identities: give each integration its own key or service account, restrict it to the APIs it needs, store it in a secrets manager rather than in front-end code, and rotate it when people leave. Billing on the paid tier also makes a leaked key a direct cost risk, since image output is the most expensive token class. The AI Model Provenance and Watermarking Guide covers how to fold provenance checks into content workflows.
Release history
| Date | Event (Gemini API changelog unless noted) |
|---|---|
| 12 March 2025 | Experimental Gemini 2.0 Flash with native image generation and editing |
| 7 May 2025 | gemini-2.0-flash-preview-image-generation released |
| 26 August 2025 | Gemini 2.5 Image Preview launched (the first Nano Banana) |
| 2 October 2025 | Gemini 2.5 Flash Image generally available |
| 4 November 2025 | Input tokens per image for Gemini 2.5 Flash Image cut from 1,290 to 258 |
| 14 November 2025 | Gemini 2.0 Flash image generation previews shut down |
| 20 November 2025 | gemini-3-pro-image-preview released (Nano Banana Pro preview) |
| 15 January 2026 | gemini-2.5-flash-image-preview shut down |
| 26 February 2026 | Nano Banana 2 preview (gemini-3.1-flash-image-preview) launched |
| 20 May 2026 | Google Pics announced for trusted testers; C2PA verification in the Gemini app (Google I/O post) |
| 28 May 2026 | Nano Banana 2 and Nano Banana Pro generally available; video-to-image added |
| 25 June 2026 | Nano Banana 2 and Pro preview IDs shut down |
| 30 June 2026 | Nano Banana 2 Lite generally available |
| 17 August 2026 | Imagen 4 models shut down, replacement gemini-nano-banana-2.1 (deprecations page) |
| 6 October 2026 | Nano Banana 2.1 generally available; Nano Banana 2 deprecated |
The cadence matters for planning: between August 2025 and October 2026 Google shipped four Nano Banana generations and retired at least five Gemini image model IDs, plus the Imagen line.
Limits
- Retirement churn: preview IDs have lived for a few months. Pin to GA IDs and watch the deprecations page.
- Mixed status signals: Google's pricing page and deprecations page gave different shutdown dates for gemini-2.5-flash-image on the check date (2 October 2026 and 15 March 2027). The deprecations page describes its dates as earliest possible; confirm against the console before planning around either.
- Feature gaps by model: Nano Banana 2 Lite has no Search grounding and 1K output only; 512px output exists only on Nano Banana 2.
- Cloud location: gemini-3-pro-image on Google Cloud is listed in the global location only, which matters for data residency reviews.
- Unsupported Cloud features: Google Cloud lists structured output, function calling, tuning, code execution and the Live API as not supported for gemini-3-pro-image.
- Licence and ownership: the Gemini API pages checked do not set out output ownership or a revenue threshold; read the Gemini API terms for commercial use.
Practitioner checklist
- Inventory every application that calls a Nano Banana model ID and record the ID, owner and Google's listed shutdown date.
- Replace any remaining preview or Imagen IDs, and plan the gemini-2.5-flash-image move to gemini-3.1-flash-lite-image before 15 March 2027.
- Confirm image workloads run on the paid API tier or Google Cloud, where Google states content is not used to improve products.
- Set a budget alert that accounts for billed thinking tokens and 4K output prices.
- Tell Gemini app users how Keep Activity affects training use and retention, and steer work images to work accounts.
- Keep generated originals so SynthID and any C2PA credentials survive, and label AI images where your disclosure rules require it.
- Issue one scoped API key or service account per integration, store it in a secrets manager and rotate it on a schedule.
- Add consent checks for edits of photos showing real people, including staff and customers.
Standards and references
- Google: Gemini API image generation (Nano Banana)
- Google: Gemini Developer API pricing
- Google: Gemini API deprecations
- Google: Gemini API release notes
- Google Cloud: Gemini 3 Pro Image (Nano Banana Pro)
- Google DeepMind: SynthID
- Google: I/O 2026 announcements
- Google: Gemini Apps limits and feature availability
- Google: Gemini Apps Privacy Hub
Related NHI Mgmt Group resources: SynthID vs C2PA Content Credentials · AI Image and Video Generation at Work · Gemini for Google Workspace Security Guide · AI Content Provenance and C2PA Guide