Version

Nano Banana Pro

Nano Banana Pro is gemini-3-pro-image: up to 4K, six object references and five character references, at $0.134 per 1K or 2K image. The problem is that Nano Banana 2, Google own cheaper and newer model, scores level or above it on both Arena boards at half the price.

  • up to 4K
  • six object and five character references
  • $0.134 per 1K image

current Maker: Google

Last checked: This is a reference page. It is re-checked against the vendor sources and updated when a new version ships. What changed

Google most expensive option, and it takes fewer votes

On the text-to-image board Nano Banana Pro scores 1246 and Nano Banana 2 scores 1264. On the editing board it is 1389 against 1385, which is a tie in practice. Now put the price next to it: Pro is $0.134 for a 1K image and Nano Banana 2 is $0.067 for the same size. Exactly double, for output people do not on average prefer.

The two boards have to be read separately, because one gap is real and the other is not. On text-to-image the distance is about eighteen points and the confidence intervals do not meet: Nano Banana 2 starts at 1258.7 and Pro ends at 1249.4. Pro is genuinely behind there. On editing Pro is four points ahead and the intervals overlap completely, which means a tie rather than a win. Put together: Pro loses one board, draws the other, and costs twice as much.

So when is Pro worth it

In two places, both of them in Google own documentation. First, 4K: the four-kilo tier is $0.24 on Pro and $0.151 on Nano Banana 2, so the higher you go the smaller the gap gets, and at the top tier it is no longer double. Second, character references: Pro takes up to five character images and Nano Banana 2 takes four. If your work is one scene with several people who have to stay themselves, that one extra image may be the only thing you need.

A reference count that runs the wrong way

Google publishes two separate numbers per model: object images and character images. Pro takes six objects and five characters, Nano Banana 2 takes ten objects and four characters, and Nano Banana 2 Lite takes fourteen objects. So the cheapest model accepts the most objects and the dearest accepts the most characters. Our table builds the reference column from the object count so the two Google models can be compared at all, and that is exactly what puts Pro low in that column. Its character number is written here because it does not fit in a table.

How many pixels is 4K

Google does not say. The image generation page writes "up to 4K" for Pro and nowhere prints pixel dimensions for any of the 1K, 2K or 4K settings. For contrast: OpenAI publishes a 3840 pixel ceiling on the long edge and a total pixel count between 655,360 and 8,294,400 for its own model. That is why the category table has no resolution column: you cannot build a column out of one label and one number.

From Iran

We read the Gemini API available-regions page today and Iran is not on it. That is the same document we read for every other Google API model, and it says nothing about the Gemini app.

Specifications

Model ID gemini-3-pro-image Source
Open weights no Source
Input text, image Source
Output image Source
Where it ships Gemini API Source
Resolution ceiling 4K Source
Price per 1K or 2K image $0.134 per image Source
Object reference images 6 reference inputs Source

Using it from Iran

This column carries a date and says how it was checked, because most listicles guess it. Where we have only read a vendor policy page, the note below says exactly that.

Reachable
blocked
Payment
no working route
Free tier
no
Measured on

How we checked: The Gemini API available-regions page lists the countries where the API works, and Iran is not on that list. That is reading a page, not measuring a network.

What it is good at

  • Up to 4K, and at the 4K tier the price gap with the cheaper model narrows
  • The most character reference images of any Google image model: five
  • Google itself describes its text rendering as advanced, for infographics and menus
  • Ships on the same API as the rest of Gemini, with no separate route to learn

Where it falls short

  • Nano Banana 2 scores level or higher on both Arena boards at half the price.
  • Google publishes no pixel dimensions for 4K, so 4K here is a label rather than a number.
  • It sits behind its own cheaper model on object references: six against ten.
  • No weights are published and there is no local route.
  • Iran is not on the Gemini API available-regions list.
  • It carries two rows with two different scores on both boards, so the score follows the serving configuration, not only the model.

Our take

If budget is not the constraint and you need several consistent characters in one scene, Pro makes sense. In every other case Nano Banana 2 does the same job for half the money, and that is what our evidence says. This is not a preference: if the leaderboard moves, this page moves with it.

Questions people actually ask

What is Nano Banana Pro

It is the marketing name for Google image model gemini-3-pro-image. It generates up to 4K, takes six object references and five character references, and ships on the ordinary Gemini API.

How much does Nano Banana Pro cost

$0.134 per 1K or 2K image and $0.24 per 4K image on the standard tier. The batch route is half of those numbers.

What is the difference between Nano Banana Pro and Nano Banana 2

Pro is gemini-3-pro-image and 2 is gemini-3.1-flash-image, so 2 is the newer one. Pro takes more character references and costs twice as much; 2 scores level or higher on both Arena boards.

Is it available from Iran

Iran is not on the Gemini API available-regions list, so there is no official route. We do not sell accounts and we do not suggest ways around it.

Sources

  1. Google, image generation in the Gemini API vendor source 12 August 2026
  2. Google, Gemini API pricing vendor source 12 August 2026
  3. Google, Gemini API available regions vendor source 12 August 2026
  4. Arena, text to image leaderboard vendor source 12 August 2026