Nano Banana Pro
Nano Banana Pro is the consumer name for Gemini 3 Pro Image, Google DeepMind's premium image generation and editing model, announced on 20 November 2025. It is the Nano Banana tier built for legible text, long instructions and work from many reference photos. It sits above Nano Banana 2 on price and below it on speed.
Anvisha Pai, Co-founder & CEO, ModaTested Google DeepMind · released
›What we asked for
A self-portrait of Nano Banana Pro, Google's premium AI image model, imagined as an editorial studio photograph: a sleek banana-yellow humanoid figure in a tailored charcoal suit, seated on a designer chair against a deep navy backdrop, magazine-cover lighting, shot on medium format film.
fal-ai/nano-banana-pro
aspect_ratio: 4:3, resolution: 1K, output_format: png, num_images: 1
returned 1200 by 896 pixels
At a glance
- Sizes
- Three sizes, 1K, 2K and 4K, in ten fixed shapes from wide 21:9 to tall 9:16. You cannot ask for an exact pixel size.
- Reference photos
- Up to 14 reference images in one request, and it keeps up to five people looking like themselves across images.
- Thinking
- It reasons before it draws and you cannot turn that off, which is part of why it is slow and good at instructions.
- Web grounding
- Can look things up on Google Search before drawing, for prompts that need current facts.
- No transparent backgrounds
- Output is always an opaque PNG, JPEG or WebP. GPT Image 2 does cut-outs; this does not.
Other Nano Banana models: Nano Banana 2, Nano Banana 2 Lite. Compare all 3.
Anvisha's take
I reach for Nano Banana Pro when the picture has words in it. Every poster line and every infographic heading came back correct on the first try, which is still not something I take for granted. The price of that is patience: a chart took 87 seconds, and it pads briefs with props and invented copy I then have to crop out. For photos and everyday posts I use Nano Banana 2 and keep Pro for the text-heavy work it is clearly built for.
Nano Banana Pro real output: the same 12 prompts every model gets
Every model on this site gets the same 12 prompts, so you can compare like with like. These are the pictures Nano Banana Pro returned, one try each, nothing re-rolled or retouched. Generated on . The 12 pictures took 15 to 87 seconds each and cost $1.80 in total.
Spelling, kerning and hierarchy of three lines of set text.
›What we asked for
A concert poster that reads 'MODA SUMMER SESSIONS' in large bold type, with the subtitle 'Friday 12 June, 8pm, Roundhouse' and a smaller line 'Tickets at the door'. Risograph style, two colours, tangerine and navy, grainy ink texture.
fal-ai/nano-banana-pro
aspect_ratio: 3:4, resolution: 1K, output_format: png, num_images: 1
returned 896 by 1200 pixels
Catalogue lighting, a clean seamless background and a believable contact shadow.
›What we asked for
Studio product photograph of a matte black ceramic pour-over coffee dripper on a seamless white background, soft top light, subtle contact shadow, centred, catalogue style.
fal-ai/nano-banana-pro
aspect_ratio: 1:1, resolution: 1K, output_format: png, num_images: 1
returned 1024 by 1024 pixels
A natural face, correct finger count and hands holding an object.
›What we asked for
Close-up portrait of a woman in her 40s with short grey hair, laughing, holding a red ceramic mug with both hands, natural window light, 85mm lens, shallow depth of field.
fal-ai/nano-banana-pro
aspect_ratio: 3:4, resolution: 1K, output_format: png, num_images: 1
returned 896 by 1200 pixels
A flat single-colour mark with a legible wordmark and no gradients or photo texture.
›What we asked for
A flat vector logo mark for a bicycle repair shop called 'Spoke & Chain': a single-colour icon combining a chain link and a wheel spoke, with the wordmark below in a geometric sans-serif, black on white.
fal-ai/nano-banana-pro
aspect_ratio: 1:1, resolution: 1K, output_format: png, num_images: 1
returned 1024 by 1024 pixels
Many distinct objects and people in one frame without merging or duplicating them.
›What we asked for
A crowded Saturday farmers market seen from slightly above: at least twelve stalls with striped awnings, shoppers carrying tote bags, a dog on a leash, crates of oranges, a musician with a guitar, bunting overhead, late morning light.
fal-ai/nano-banana-pro
aspect_ratio: 4:3, resolution: 1K, output_format: png, num_images: 1
returned 1200 by 896 pixels
A named historical style applied faithfully rather than a generic painterly look.
›What we asked for
A lighthouse on a rocky cliff during a storm, in the style of a Japanese ukiyo-e woodblock print, with visible wood grain texture, flat colour areas and a bokashi gradient sky.
fal-ai/nano-banana-pro
aspect_ratio: 3:4, resolution: 1K, output_format: png, num_images: 1
returned 896 by 1200 pixels
Six labelled bars at the stated heights, in order, with readable axes.
›What we asked for
A clean bar chart titled 'Monthly active users' showing six months, January through June, with the values 120, 150, 180, 170, 220 and 260, labelled axes, white background, one accent colour, no 3D effects.
fal-ai/nano-banana-pro
aspect_ratio: 4:3, resolution: 1K, output_format: png, num_images: 1
returned 1200 by 896 pixels
A 21:9 composition that leaves the left third clear for headline text.
›What we asked for
Website hero banner: a wide panorama of a sunlit modern co-working space with plants, the left third of the frame kept as clear wall space for text, warm neutral palette.
fal-ai/nano-banana-pro
aspect_ratio: 21:9, resolution: 1K, output_format: png, num_images: 1
returned 1584 by 672 pixels
Four numbered steps in the given order with their exact wording, plus an icon each.
›What we asked for
An infographic card titled 'How to brew pour-over' with four numbered steps in this order: 1. Rinse the filter, 2. Bloom for 30 seconds, 3. Pour in slow circles, 4. Wait 3 minutes. A simple icon beside each step, muted palette, clean sans-serif type.
fal-ai/nano-banana-pro
aspect_ratio: 3:4, resolution: 1K, output_format: png, num_images: 1
returned 896 by 1200 pixels
Appetising, menu-grade food with correct textures and no melted anatomy.
›What we asked for
Overhead photograph of a margherita pizza on a wooden board, fresh basil leaves, one slice pulled out with the cheese stretching, restaurant menu photography, natural light.
fal-ai/nano-banana-pro
aspect_ratio: 1:1, resolution: 1K, output_format: png, num_images: 1
returned 1024 by 1024 pixels
Exact counts and left-right placement of five named objects.
›What we asked for
A glass of water on a wooden table with exactly three ice cubes in it, a yellow pencil resting across the rim of the glass, a blue book to the left of the glass and a red apple to the right, photographed from table height.
fal-ai/nano-banana-pro
aspect_ratio: 4:3, resolution: 1K, output_format: png, num_images: 1
returned 1200 by 896 pixels
Change one object's colour and add a part while leaving background, light and framing untouched.
›What we asked for
Change the coffee dripper to a deep forest green and add a small brass handle on its right side. Keep the background, lighting and composition exactly as they are. (applied to the product on white result above)
fal-ai/nano-banana-pro/edit
aspect_ratio: 1:1, resolution: 1K, output_format: png, num_images: 1
returned 1024 by 1024 pixels
How the cost under each picture was worked out: fal unit price, 0.15 USD per image, read from fal's pricing API at run time.
Nano Banana Pro pricing
| Where | Unit | Price | Source |
|---|---|---|---|
| Gemini API | 1K or 2K image Google states the per-image figure itself; billing is by token underneath. | $0.134 | Gemini API pricing, read |
| Gemini API | 4K image | $0.24 | Gemini API pricing, read |
| Gemini API, Batch | 1K or 2K image Batch halves the rate; 4K in batch is $0.12. | $0.067 | Gemini API pricing, read |
| fal | 1K or 2K image Flat per-image price on the generate and edit endpoints. This is what our sample run was billed at. | $0.15 | fal model page and pricing API, read |
| fal | 4K image fal charges 4K at double the standard rate. | $0.30 | fal model page, read |
Is Nano Banana Pro free?
There is no free way to generate with Nano Banana Pro today. The Gemini API has no free tier for any Nano Banana model, and in the Gemini app free accounts generate with Nano Banana 2; the option to redo an image with Nano Banana Pro is reserved for Google AI Plus, Pro and Ultra plans.
- •Gemini API: Google's pricing page lists the free tier as not available for gemini-3-pro-image.
- •Gemini app, free: images generate with Nano Banana 2 (or 2 Lite depending on the model setting) and download at 1K.
- •Gemini app, Google AI Plus, Pro and Ultra: can redo an image with Nano Banana Pro and download at 2K. Google publishes no daily number and says limits may change without notice.
- •At launch in November 2025 free accounts had a small Nano Banana Pro quota before falling back to Nano Banana; the current help pages no longer describe one.
- •Every image carries an invisible SynthID watermark. At launch Google said free and Pro tiers also get a visible Gemini sparkle mark and Ultra does not.
Where to use Nano Banana Pro
- Gemini app · Choose Create images with the Thinking model. Pro redo needs a paid plan.
- Google AI Studio and Gemini API · Model id gemini-3-pro-image. Pay per token, no free tier.
- Vertex AI · Same model for Google Cloud accounts. Some controls differ; see quirks.
- fal · Flat $0.15 per image, generate and edit endpoints.
- Replicate · Hosted endpoint under google/nano-banana-pro.
- Moda · Available as the NanoBanana Pro image model inside the Moda design agent, with 1K, 2K and 4K output.
Generate with Nano Banana Pro in Moda
Pick Nano Banana Pro in the Moda design agent, then edit the result as vectors, text and layers rather than a flat image.
Open Moda →What we noticed running Nano Banana Pro
- •Set text is its strength. All three lines of the poster came back spelled correctly with sensible hierarchy on the first try, and the four numbered steps of the infographic kept their exact wording and order. See the text in an image picture.
- •It elaborates beyond the brief. The infographic was asked for four headings and got four headings plus a paragraph of invented instructions under each. Say 'no other text' when you mean it. See the infographic with steps picture.
- •It adds props you did not ask for, and small incidental text degrades. The pizza shot arrived with a hand, a film border, a bowl of chilli flakes and a menu card whose fine print is misspelled, while the large text elsewhere on the page was clean. See the food photography picture.
- •Style prompts come with fabricated signatures and seals. The ukiyo-e print carries invented Japanese characters, a red artist's seal and a signature in the margin. Crop or repaint them before commercial use. See the specific art style picture.
- •Negative-space instructions are taken literally. Asked to keep the left third clear for text, it rendered a flat blank wall with a hard vertical edge, which reads as a composite rather than a photograph with room in it. See the wide aspect ratio picture.
- •Left-right placement was right and the pencil balanced on the rim, but the count was not: the glass holds five or six ice cubes, not three. Exact counts of small things are still a weak spot. See the hard composition picture.
- •Charts come out looking like a default plotting library, correct values and all. Every bar was the right height and in order, and it also took 87 seconds, the slowest of the twelve; the model reasons before it draws and a chart makes it reason a lot. See the chart picture.
- •Latency is wide. Across the twelve generations at 1K, the fastest returned in 15 seconds and the slowest in 87, with most between 15 and 40. Budget for that in anything interactive.
- •1K is not 1024 on every side. At 3:4 the output was 896 by 1200, at 4:3 1200 by 896, at 21:9 1584 by 672 and at 1:1 1024 by 1024. Ask for 2K or 4K if you need a long edge over about 1600 pixels.
- •On Vertex AI the thinking level cannot be set for this model: both MINIMAL and HIGH return a 400 error, while the same request is accepted through the Gemini API with an AI Studio key. The two backends disagree, so test on the one you deploy to.
- •Edits keep the brief. Recolouring the dripper and adding a brass handle left the background, shadow and framing intact, though the rim and body were subtly resculpted rather than repainted. See the edit an existing image picture.
Nano Banana Pro limits and API parameters
| Limit | Value |
|---|---|
| Output size | 1K, 2K or 4K in ten fixed shapes. Our 1K run returned 1024 by 1024 at square, 896 by 1200 at 3:4 and 1584 by 672 at 21:9. |
| Reference images | Up to 14 in total, with consistency for up to 5 people. |
| Transparent background | Not supported. fal API reference, read |
| Rate limits | Google meters images per minute per project tier and shows the live numbers in AI Studio rather than publishing a table. Gemini API rate limits, read |
›Every setting, for developers
fal (fal-ai/nano-banana-pro)
fal API reference, read
| Parameter | Type | Values | Default |
|---|---|---|---|
| prompt | string | Any text Required. | |
| aspect_ratio | enum | auto, 21:9, 16:9, 3:2, 4:3, 5:4, 1:1, 4:5, 3:4, 2:3, 9:16 | auto |
| resolution | enum | 1K, 2K, 4K | 1K |
| num_images | integer | 1 or more | 1 |
| seed | integer | Any integer Omit for a random seed. | |
| output_format | enum | jpeg, png, webp | png |
| safety_tolerance | enum | 1 to 6, 1 is strictest | 4 |
| system_prompt | string | Any text | empty |
| enable_web_search | boolean | true, false Google Search grounding. | false |
| limit_generations | boolean | true, false | true |
| image_urls | list of URLs | Reference images Edit endpoint only, where it is required. |
Gemini API (gemini-3-pro-image)
Gemini API image generation guide, read
| Parameter | Type | Values | Default |
|---|---|---|---|
| aspectRatio | enum | 1:1, 3:2, 2:3, 3:4, 4:3, 4:5, 5:4, 9:16, 16:9, 21:9 | |
| imageSize | enum | 1K, 2K, 4K | |
| reference images | inline images | Up to 6 objects, 5 characters and 3 style references in one request | |
| thinking | built in | Always on for this model On Vertex AI the level cannot be set; see quirks. | |
| grounding | tool | Google Search (web) |
What Nano Banana Pro will not draw
Google's Generative AI Prohibited Use Policy applies on every surface. It bars child sexual abuse material, non-consensual intimate imagery, content that facilitates violent extremism, self-harm or illegal activity, hate speech and harassment, and impersonating a real person without disclosure in order to deceive. There is no adult content mode. fal exposes a safety_tolerance dial from 1 to 6 but it operates inside the same policy. Every image is watermarked with SynthID.
Alternatives to Nano Banana Pro
Frequently asked questions
What is Nano Banana Pro?
Nano Banana Pro is Google's name for Gemini 3 Pro Image, its premium image generation and editing model, released on 20 November 2025. It is the Nano Banana tier built for legible text, complex instructions and 4K output.
How much does Nano Banana Pro cost?
On the Gemini API a 1K or 2K image is $0.134 and a 4K image is $0.24, read on 9 September 2026. On fal it is a flat $0.15 per image at 1K or 2K and $0.30 at 4K.
Is Nano Banana Pro free?
No. The Gemini API has no free tier for it, and free Gemini app accounts generate with Nano Banana 2 instead. Redoing an image with Nano Banana Pro needs a Google AI Plus, Pro or Ultra plan.
What is the difference between Nano Banana Pro and Nano Banana 2?
Nano Banana 2 is the faster, cheaper generalist that became the Gemini app default in February 2026. Nano Banana Pro is the slower premium tier for the hardest briefs. On our fixed prompt set the gap shows most on dense text and busy scenes.
Does Nano Banana Pro have an API?
Yes. The model id is gemini-3-pro-image on the Gemini API and Vertex AI, and it is hosted on fal as fal-ai/nano-banana-pro and on Replicate.
Can Nano Banana Pro make transparent backgrounds?
No. It returns opaque images only. GPT Image 2 supports a transparent background setting if you need cut-outs.
Is Nano Banana Pro actually good?
On our fixed prompt set, yes for text and layouts: every line of the poster and every heading of the infographic came back correct. It is weaker on exact counts of small objects, it adds props and captions you did not ask for, and it is the slowest of the three Nano Banana models, with generations taking 15 to 87 seconds at 1K.
How this page is made
Prices, limits and settings are read from the linked vendor and host pages on the dates shown. Every picture was generated by us through fal with the request shown under it, and the original files are kept unedited. A generation that failed is shown as a failure. Model and vendor names are used to identify the products; no vendor imagery appears on this page.
