Minute Verdict
Nano Banana is Google’s name for the image models built into Gemini, billed per picture on the Gemini API from $0.0336 for a 1K image, against $0.134 on Nano Banana Pro. A thousand 1K pictures at the cheap rate runs $33.60, and the Gemini app folds limited access into its plans instead of metering it.
Best for: a developer or small studio that needs generation and conversational editing through one API key and can watch a per-image meter. Not for: a team that wants a flat monthly price per seat, or a free API tier to prototype on. From: $0.0336 per 1K image, checked November 4, 2026
What Nano Banana does
Nano Banana is the product name for Gemini’s native image generation. The model can take text, images, video or a mix of them in one request, produce a picture, then edit that picture when you describe what should change. The published model ids keep the technical name: gemini-nano-banana-2.1, gemini-3.1-flash-image, gemini-3.1-flash-lite-image, gemini-3-pro-image and the legacy gemini-2.5-flash-image.
Five models sit under the name, and they are not interchangeable. Nano Banana 2 Lite is the cheapest and fastest, built for volume, and Google states plainly that it is not optimized for several reference inputs or for sequential multi-turn editing. Nano Banana 2.1 is the current workhorse, an update to Nano Banana 2 with better text rendering, multi-turn consistency and Google Search grounding. Nano Banana 2 is the previous workhorse, still the one that generates 4K comfortably and holds several references. Nano Banana Pro is the premium model: it spends thinking tokens on the prompt first, then draws, which is what brand consistency and long text inside an image need.
The feature list matters more than the model names. Output runs at 1K, 2K and 4K, with a 512px step exclusive to Nano Banana 2. The 2.1 and 2 models accept up to 14 reference images in one call, add the tall 1:4 and 1:8 ratios and the wide 4:1 and 8:1, and can call Google Search, or Google Image Search, before drawing, which is how a chart with today’s numbers gets generated. Every output carries a SynthID watermark.
On this page
What you get for the money
| Plan | Price | What is included | The limit that matters |
|---|---|---|---|
| Nano Banana 2 Lite | $0.0336 per 1K image, $0.0168 in batch | The fastest image model, 1K output, text and thinking at $1.50 per million tokens | Not built for several reference images or for sequential editing |
| Nano Banana 2.1 | $0.0336 per 1K image, $0.0504 at 2K, $0.113 at 4K | Search and Image Search grounding, up to 14 references, the new tall and wide ratios | Input is billed at $1.50 per million tokens |
| Nano Banana 2 | $0.045 at 512px, $0.067 at 1K, $0.101 at 2K, $0.151 at 4K | 4K generation and several references, the model 2.1 updates | Superseded for new projects by 2.1 |
| Nano Banana Pro | $0.134 per 1K or 2K image, $0.24 at 4K | Thinking mode, brand consistency, high-fidelity text inside the image | Four times the 2.1 rate at 1K, and no free tier |
| Nano Banana, legacy | $0.039 per image | The original 2.5 Flash Image model | Retired on March 15, 2027, with a migration path to 2 Lite |
| Gemini app, free plan | $0 | Image generation and editing inside the Gemini app | Access level tied to the plan, no published image count |
| Google AI Pro | $19.99 a month | More access to Nano Banana in Search, limited Nano Banana Pro inside Flow | Not an image quota |
| Google AI Ultra | $199.99 a month | The highest access to Nano Banana across Google surfaces | The most expensive route to the same models |
The meter is what this product is, and it is legible. A thousand 1K images cost $33.60 on 2.1 or 2 Lite, and the same thousand at 4K on Pro cost $240, which is the whole argument for choosing a model per job rather than per habit. Batch work halves the 2.1 rate to $0.0168 per image, so a large library built in one overnight job costs $16.80 per thousand rather than $33.60.
Two charges sit outside the image table. Input tokens are billed on top, at $1.50 per million on 2.1, $0.50 on 2, $0.25 on 2 Lite and $2.00 on Pro, and an input image on Pro is priced at $0.0011. Grounding with Google Search gives 5,000 free search requests a month across Gemini 3 and newer models, then bills $14 per 1,000, with a daily cap of 1,500 requests.
Where it falls short
The API has no free tier for images. The pricing table marks the free column as not available for all five models, so a developer who wants to evaluate the quality against a real brief has to attach billing first. The free door is the Gemini app, and it is an access level, not a documented count.
Resolution drives the bill more than volume does. A 4K image on Pro costs $0.24 against $0.134 at 1K, nearly double, and a 4K image on 2.1 costs three times its 1K price. A workflow that assumes “an image is an image” will be wrong by a factor of three.
The name now covers four prices. A buyer who reads “Nano Banana” on a wrapper cannot tell whether the call will cost $0.0336, $0.067 or $0.134 per picture, and the difference is the model id in the request.
Model churn is a maintenance line. The legacy model that made the name famous retires on March 15, 2027, with 2 Lite named as the migration target, and the documentation recommends 2.1 for new projects while 2 is still the one described for 4K work.
Finally, the watermark is not optional. Every generated image carries SynthID, and no opt-out is described on the image generation page. For a buyer who needs a clean asset, that is a constraint to know before the first invoice.
Nano Banana alternatives
These are the tools buyers compare it against, priced on the same day.
- Midjourney, from $10 a month for the Basic plan with 200 minutes of fast GPU time, the pick when the look matters more than the API: our Midjourney review.
- Leonardo AI, $12 a month for Essential with 8,500 fast tokens and access to more than 80 models, the option when you want a studio interface rather than an endpoint: our Leonardo AI review.
- Claid AI, $9 a month, built for product photos rather than free scenes, the closest thing we have reviewed to the editing loop Nano Banana sells: our Claid AI review.
Who it is for: Nano Banana
It fits a builder. If you already have a product that generates or edits pictures, the API is the shortest path to a Google image model: one key from Google AI Studio, a model id, a per-image rate you can put in a spreadsheet, and an editing call that takes the previous picture as input. It also fits a small studio that batches a catalogue overnight, since the pair of 2.1 at $0.0336 and the batch rate at $0.0168 keeps a thousand images under $34.
It does not fit a buyer who wants a fixed price per person. There is no seat here, no cap, and no invoice that can be predicted without knowing how many pictures will be made and at what size. It does not fit a beginner who wants an app to click through either: the Gemini app covers that, but the models reviewed here are reached through a key, and the documentation assumes a request body.
For a team that needs images inside a document workflow rather than in a product, start with the app plan at $0 and move to the API only once the access level is the limit.
How we checked
This is a research review. We have not used Nano Banana, so there is no sample image, no prompt we ran and no comparison we drew ourselves.
Every rate on this page was read on November 4, 2026, on the Gemini developer pricing page, the image generation documentation, the deprecation table and the Gemini subscription page. The per-image figures, the batch rate, the free tier column, the resolution steps, the 14 reference images, the SynthID policy and the March 15, 2027 retirement date all come from those four pages. What a test would settle is what no document can: how the 2.1 output compares with Pro on the same prompt. There is no affiliate program to disclose, and our editorial policy covers how the site is funded.
What to do next
Start with the model, not the plan. Price one real job at 1K on 2.1 and the same job at 4K on Pro before you build anything, because the ratio between them decides your architecture. If the images need text, a brand palette or a specific person across several turns, budget for Pro at $0.134 and treat 4K as a separate line. Then check the model id in your code against the deprecation table, since the legacy name is on a clock.
FAQ
Is Nano Banana free? In the Gemini app the free plan at $0 includes image generation and editing. The Gemini API is a different meter, and its free tier is marked as not available for every Nano Banana model.
How much does one image cost? $0.0336 per 1K image on Nano Banana 2.1 and on Nano Banana 2 Lite, $0.067 on Nano Banana 2, and $0.134 on Nano Banana Pro. The 4K output costs $0.113 on 2.1, $0.151 on 2 and $0.24 on Pro. Batch work halves the 2.1 rate to $0.0168.
What is the difference between Nano Banana, Nano Banana 2, 2.1 and Pro? They are four members of one family. Nano Banana is the legacy 2.5 Flash Image model, Nano Banana 2 is the previous high-efficiency model, 2.1 is its update with search grounding and better text, and Pro is the premium model that spends thinking tokens before it draws.
Can Nano Banana make video? The image generation documentation describes images only. Video sits outside this family, and nothing on those pages presents the models as a video tool.
Do I need a Google Cloud account? An API key from Google AI Studio is enough to call the models, and the same pricing page carries the rates. Google AI Studio itself is free.
Sources
Sources
- https://ai.google.dev/pricing (checked November 4, 2026)
- https://ai.google.dev/gemini-api/docs/image-generation (checked November 4, 2026)
- https://ai.google.dev/gemini-api/docs/deprecations (checked November 4, 2026)
- https://gemini.google/subscriptions/ (checked November 4, 2026)
Update log
Published November 4, 2026. Every rate and limit checked November 4, 2026 against the Gemini developer pricing page, the image generation documentation, the deprecation table and the Gemini subscription page.
