Nano Banana API: Pricing, API Keys, and Which Model to Actually Use

If you have looked at the Nano Banana model list and felt stuck, that reaction is correct. There are nine variants. They are named nano-banana, nano-banana-2, nano-banana-pro, and then each of the latter two again at 1k, 2k and 4k. Nothing in the names tells you that the gap between the cheapest and most expensive is roughly 99x.

This is a guide to the three questions people actually ask: what it costs, how to get a key, and which one to pick.

How the pricing is structured

Nano Banana bills per call, not per token. Every generation costs the same whether your prompt is five words or five hundred. That is unlike a text model and it changes how you should think about cost: prompt engineering is free, retries are not.

Three things drive the price.

The tier. nano-banana-2 is the workhorse. nano-banana-pro is the high-quality tier, and at the same resolution it costs about 25x more. That is not a typo, and it is the single most important number in this article. If you are generating at Pro when 2 would have done, you are paying twenty-five times too much.

The resolution. Each step up roughly multiplies the cost:

Tier	1k → 2k	2k → 4k
Nano Banana 2	~1.5x	~1.5x
Nano Banana Pro	~2x	~2x

Pro scales more steeply with resolution than 2 does, so nano-banana-pro-4k is where cost concentrates: it is roughly 99x the price of nano-banana-2-1k.

Your account group. The rate you pay is your account's group rate applied to the model's base price. Two accounts calling the same model can pay different amounts, which is why this article gives you ratios rather than dollar figures. For the number that applies to you, read the pricing page — it reflects your account when you are signed in, and it is the only figure that will not go stale.

The original nano-banana is still available and sits slightly below nano-banana-2-1k. It is the older model; unless you have a reason to pin it, start with 2.

Getting an API key

1. Sign up at ccapi.ai
2. Open Console → Tokens and click Create token
3. Name it something you will recognise, leave the rest as-is, and submit
4. Copy the key — it starts with sk- and is shown once

That key works for every model on the platform, not just Nano Banana, and for both the REST API and the MCP integration. There is no separate Google Cloud project to set up and no Gemini key to manage, which is usually the reason people go looking for an aggregator in the first place.

Add a little credit under Console → Top Up. Because billing is per call, you can estimate a batch exactly: number of images times the per-call rate for the model you picked.

Which model should you use

Work from the job, not the name.

Drafting, iterating, or anything a human will look at once. Use nano-banana-2-1k. It is the cheapest of the current tier and the difference is invisible at thumbnail or in-app sizes. Most people should be here for most calls.

Production assets at normal display size. nano-banana-2-2k. About 1.5x the 1k cost, and the extra pixels matter once an image is a hero or a card that someone looks at directly.

Print, large displays, or anything that will be cropped hard. nano-banana-2-4k. Beyond this, ask whether you need Pro or just need pixels — they are different problems, and 4k on the 2 tier is far cheaper than 1k on Pro.

When quality is the constraint, not resolution. nano-banana-pro-1k. This is the case people most often get wrong in the other direction: if the 2 tier is producing the wrong kind of image — composition, prompt adherence, text rendering — more resolution will not fix it, and Pro might. Test at 1k before committing to Pro at higher resolutions.

nano-banana-pro-4k. Only when you know why. At roughly 99x the cheapest option, this should be a deliberate decision with a reason attached, not a default someone set once and forgot.

A practical pattern: draft on nano-banana-2-1k until the prompt is right, then regenerate the final at whatever tier the output actually needs. Because prompt length is free, the iteration costs you almost nothing.

The full variant list

All nine, with the job each one is for. Prices are shown as multiples of nano-banana-2-1k so they stay accurate regardless of your account rate.

Model	Relative cost	Use it for
nano-banana	~0.75x	The previous generation. Still live; pin it only if you have a reason.
nano-banana-2-1k	1x (baseline)	Drafting, iteration, in-app and thumbnail sizes. Start here.
nano-banana-2-2k	~1.5x	Production assets at normal display size.
nano-banana-2-4k	~2.3x	Print, large displays, hard crops.
nano-banana-2	—	Alias for the 2 tier without a pinned resolution.
nano-banana-pro-1k	~25x	Quality problems that resolution cannot fix. Test here first.
nano-banana-pro-2k	~49x	Pro output at display size.
nano-banana-pro-4k	~99x	Deliberate decisions only.
nano-banana-pro	—	Alias for the Pro tier without a pinned resolution.

The two unsuffixed aliases exist so you can write nano-banana-pro and let the platform apply its default resolution. Pin the resolution explicitly in production — the default can change, your invoice should not.

Calling it

The endpoint is OpenAI-compatible, so most existing image code works with a base URL and model name change:

curl https://api.ccapi.ai/v1/images/generations \
  -H "Authorization: Bearer sk-YOUR-KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "nano-banana-2-1k",
    "prompt": "a red bicycle against a whitewashed wall, morning light"
  }'

The response carries a URL. Those are re-hosted rather than pointed at the provider's storage, so links stay stable and do not expire out from under you — but download anything you intend to keep indefinitely.

To do the same thing from Claude Code, Cursor or Cline without writing a request at all, connect the MCP server and ask in plain language. The same key works.

Frequently asked questions

What is the difference between Nano Banana 2 and Nano Banana Pro? Pro is the higher-quality tier and costs roughly 25x more at the same resolution. Use 2 unless the 2 tier is failing on quality rather than on resolution.

Does prompt length affect the price? No. Billing is per call, so a long prompt costs exactly what a short one does. Retries do cost, so a failed generation is worth reading rather than repeating.

Do I need a Google or Gemini API key? No. One CCAPI key reaches Nano Banana along with every other model on the platform. There is no separate cloud project to configure.

How do I know what a call will cost before I make it? The pricing page shows the rate that applies to your account. Multiply by the number of images; per-call billing makes batch estimates exact rather than approximate.

Is the original nano-banana model still usable? Yes, and it is slightly cheaper than nano-banana-2-1k. It is the older generation, so start with 2 unless you have a specific reason to pin the old one.

Can I generate at 4k on the cheaper tier? Yes — nano-banana-2-4k exists and costs far less than any Pro variant. Resolution and quality are separate axes, and conflating them is the most common way people overspend here.

The short version

Start on nano-banana-2-1k. Move up in resolution when the output is too small, and move to Pro only when the output is wrong in a way resolution cannot fix. Check your actual rate on the pricing page, and see the Nano Banana model page for the full variant list and parameters.