Nano Banana 2 is free for casual use and cheap for developers. In the Gemini app you pay nothing beyond daily limits; on the API you pay per image, from about $0.045 at 0.5K to $0.151 at 4K — and batch mode roughly halves both. Below is the full breakdown, including the cheaper Lite model and how Google bills image output as tokens.
The short answer
There are three ways to pay for Nano Banana 2 (Gemini 3.1 Flash Image). Use it free inside Gemini, subscribe to a Google AI plan to raise the limits, or call the API per image. Most people never need to pay; developers and high-volume users go straight to the API.
Consumer plans
Inside the Gemini app and Google Search, Nano Banana 2 is free with daily generation caps. If you hit those regularly, a Google AI subscription lifts them and bundles the rest of Gemini — the same plans that power our Gemini pricing guide.
| Plan | Price | What you get |
|---|---|---|
| Free (Gemini app) | $0 | Nano Banana 2 with daily image limits, SynthID watermark |
| Google AI Pro | $19.99/mo | Higher image limits, full Gemini, 2 TB storage |
| Google AI Ultra | $99.99/mo | Top limits, highest-priority access, most storage |
For a solo creator, the free tier or Pro at $19.99/mo covers almost everything, and Pro pays for itself if you also use Gemini for writing, research, and video.
API pricing by resolution
The API is where the numbers matter. Google prices per image by resolution, and if your workload can tolerate a queue, batch mode cuts the rate by roughly half.
| Resolution | Standard | Batch mode |
|---|---|---|
| 0.5K | $0.045 | $0.022 |
| 1K | $0.067 | $0.034 |
| 2K | ~$0.10 | $0.050 |
| 4K | $0.151 | $0.076 |
So a thousand 1K images cost about $67 at standard rates, or roughly $34 in batch. That's competitive for hosted, top-tier quality, and the batch price is genuinely aggressive.
Nano Banana 2 Lite
If you want maximum throughput at minimum cost, Nano Banana 2 Lite (Gemini 3.1 Flash Lite Image) delivers a 1K image in about four seconds for roughly $0.034 — about half the standard model's 1K rate. Quality is a step below the full model, but for thumbnails, variations, and drafts where "good and instant" beats "perfect," Lite is the value pick.
The token math
Under the hood the API bills image output as tokens, which matters if you're mixing text and image calls. The standard model is $0.50 per million input tokens and $3 per million output tokens; Lite is $0.25 / $1.50. Google also publishes the flat per-image rates above, so you can budget either way — most teams just use the per-image numbers.
Is it worth it?
For casual use, there's nothing to weigh — it's free and it's the best model. For builders, the question is standard rate versus batch and full versus Lite. If latency isn't critical, batch mode plus Lite gets you high-quality images for around two cents each, which undercuts most hosted rivals at this quality tier.
The main alternative on cost is self-hosting an open model like Flux 2, which can be cheaper per image but adds GPU and ops overhead. If you'd rather not run infrastructure, Nano Banana 2's API is the easy call. For the model itself, read our full Nano Banana 2 review, or see how it compares in Nano Banana 2 vs Flux 2.
Frequently Asked Questions
How much does Nano Banana 2 cost?
It's free in the Gemini app and Google Search with daily limits. On the API it runs from $0.045 per image at 0.5K to $0.151 at 4K, with batch mode roughly halving those rates. Lite is about $0.034 per 1K image.
Is there a free version?
Yes. Nano Banana 2 is free inside the Gemini app and Google Search, subject to daily generation caps. Google AI Pro and Ultra raise those limits.
What is batch mode?
Batch mode processes API requests asynchronously in exchange for a lower price — roughly half the standard per-image rate. It's ideal when you don't need images back instantly.
Is Nano Banana 2 cheaper than Flux or GPT Image?
At standard rates it's competitive but not the cheapest — self-hosted Flux can undercut it. In batch mode at about $0.022 per 1K image, it's one of the cheapest hosted options for its quality.
Does the API charge per token?
Yes. Image output is billed as tokens at $0.50/M input and $3/M output for the standard model, and $0.25/$1.50 for Lite. Google also publishes flat per-image rates by resolution.
Do I own images I generate?
Google grants usage rights to images you create, but every output carries an invisible SynthID watermark marking it as AI-generated. Always check the current terms for commercial projects.