In short (October 23, 2026): On the Gemini API, Nano Banana Pro (gemini-3-pro-image) costs $0.134 per image at 1K or 2K and $0.24 at 4K, because generated images are billed as output tokens at $120 per million. Batch and Flex halve that, to $0.067 and $0.12. No Nano Banana model has a free tier on the API. The newer Nano Banana 2.1 costs $0.0336 per 1K image, a quarter of Pro, $0.0504 at 2K and $0.113 at 4K. Nano Banana 2 costs $0.067, $0.101 and $0.151, and Nano Banana 2 Lite $0.0336 at 1K. The original Nano Banana, gemini-2.5-flash-image, is listed as deprecated with an October 2, 2026 shutdown date.
Google prices its image models in tokens, not images, and the number of tokens an image uses depends on the model and the resolution you ask for. That is why the same 4K picture costs $0.113 on one Nano Banana model and $0.24 on another, and why on Nano Banana Pro a 2K image costs exactly as much as a 1K one. This page turns the token prices into prices per image and per 1,000 images, and shows the cheapest way to run each model.
Cost per image, every Nano Banana model
| Model (API name) | 0.5K | 1K | 2K | 4K |
|---|---|---|---|---|
Nano Banana Pro (gemini-3-pro-image) | Not listed | $0.134 | $0.134 | $0.24 |
Nano Banana 2.1 (gemini-nano-banana-2.1) | Not listed | $0.0336 | $0.0504 | $0.113 |
Nano Banana 2 (gemini-3.1-flash-image) | $0.045 | $0.067 | $0.101 | $0.151 |
Nano Banana 2 Lite (gemini-3.1-flash-lite-image) | Not listed | $0.0336 | Not listed | Not listed |
Standard tier, USD per generated image, from the Gemini API pricing page (last updated October 22, 2026). 1K is 1024x1024, 2K is 2048x2048, 4K is 4096x4096.
Why images are priced in tokens
Each model has an output price per million image tokens and a fixed token count per image size. Multiply the two and you get the per-image price above:
| Model | Image output, per 1M tokens | Tokens per image |
|---|---|---|
| Nano Banana Pro | $120 | 1,120 at 1K and 2K, 2,000 at 4K |
| Nano Banana 2.1 | $30 | 1,120 at 1K, 1,680 at 2K, 3,780 at 4K |
| Nano Banana 2 | $60 | 747 at 0.5K, 1,120 at 1K, 1,680 at 2K, 2,520 at 4K |
| Nano Banana 2 Lite | $30 | 1,120 at 1K |
Two consequences follow. On Nano Banana Pro, 1K and 2K both use 1,120 tokens, so a 2K image costs the same $0.134 as a 1K one. And Nano Banana 2.1 spends more tokens on a 4K image than Nano Banana 2 does (3,780 against 2,520), but at half the token price it still comes out cheaper, $0.113 against $0.151.
The rest of the request is billed as well. Nano Banana Pro charges $2.00 per million input tokens and counts each input image as 560 tokens, or $0.0011 per image; 2.1 charges $1.50 per million for text, image and video input, Nano Banana 2 $0.50 and Lite $0.25. Text output and thinking tokens are billed at the text output rate: $12.00 per million on Pro, $7.50 on 2.1, $3.00 on Nano Banana 2 and $1.50 on Lite. For a prompt of a few hundred words the input is a fraction of a cent (200 tokens on Pro is $0.0004), so the image itself dominates the bill unless the model writes a long reasoning trace first.
What 1,000 images cost
| Model and size | Standard | Batch |
|---|---|---|
| Nano Banana Pro, 1K or 2K | $134.00 | $67.00 |
| Nano Banana Pro, 4K | $240.00 | $120.00 |
| Nano Banana 2.1, 1K | $33.60 | $16.80 |
| Nano Banana 2.1, 2K | $50.40 | $25.20 |
| Nano Banana 2.1, 4K | $113.00 | $56.70 |
| Nano Banana 2, 1K | $67.00 | $34.00 |
| Nano Banana 2, 4K | $151.00 | $76.00 |
| Nano Banana 2 Lite, 1K | $33.60 | $16.80 |
Per-image prices from the pricing page multiplied by 1,000; image output only.
Batch, Flex and Priority on Nano Banana Pro
Of the current models, Nano Banana Pro is the only one listed with four service tiers; 2.1, 2 and Lite list Standard and Batch.
- Standard: $0.134 per 1K or 2K image, $0.24 per 4K image.
- Batch: half price: $1.00 per million text input tokens, $0.0006 per input image, $6.00 per million text and thinking output tokens, $0.067 per 1K or 2K image and $0.12 per 4K image.
- Flex: the same prices as Batch.
- Priority: $3.60 per million input tokens, $21.60 for text and thinking output, and $216 per million image tokens. With the token counts above that is about $0.242 per 1K or 2K image and $0.432 per 4K image, 80% more than Standard.
Is there a free tier?
No. The free tier column reads "Not available" for every Nano Banana model on the Gemini API pricing page, so every image generated through the API is billed from the first one. On the paid tier, Google lists your prompts and responses as not used to improve its products. Grounding with Google Search is the one free allowance in the picture: 5,000 search requests a month, shared across all Gemini 3 and newer models, then $14 per 1,000. If you are working out which Gemini features still have a free tier at all, see Gemini API billing and free tier confusion.
The cheapest way to use Nano Banana Pro
- Send anything that can wait through Batch or Flex. It halves the price of every image and every token.
- Ask for 2K instead of 1K. On Pro both cost $0.134, so the larger image is free resolution.
- Keep 4K for final assets. On Pro, 4K costs 79% more than 2K.
- Draft on Nano Banana 2.1, finish on Pro. At 1K, 2.1 costs a quarter of Pro. Google describes it as an update to Nano Banana 2 with improved visual quality, multi-turn character consistency, accurate text rendering and search-grounded generation across 1K, 2K and 4K.
- Watch the thinking tokens. They are billed at the text output rate, $12.00 per million on Pro, and do not show up in a per-image estimate.
The original Nano Banana is gone
The pricing page lists gemini-2.5-flash-image, the first model sold as Nano Banana, as deprecated with a shutdown date of October 2, 2026, and tells users to move to Gemini 3.1 Flash Image or Gemini 3.1 Flash Lite Image. Its last listed price was $0.039 per image up to 1024x1024 (1,290 tokens at $30 per million). Moving a 1K workload to Nano Banana 2 raises the price per image by about 72%, to $0.067. Moving it to Nano Banana 2 Lite or 2.1 lowers it by about 14%, to $0.0336.
Metering image generation
Because the bill is per token and the product is an image, cost per image is a function of model, size and tier, and it can move by an order of magnitude without anyone changing a line of code: a switch from 1K to 4K on Nano Banana 2.1 multiplies the price per image by 3.4, and a switch from 2.1 to Pro multiplies it by 4 at 1K. Record model, resolution and tier on every request, and compare per-customer spend against the per-image table above rather than against a token total. Our Gemini API spend caps and tiers guide covers the account-level limits that apply to image models too, and Gemini 3.8 Flash pricing covers the text model most image pipelines pair with. Day-to-day price changes are logged in our AI price tracker.
Sources
Read on October 23, 2026: the Gemini API pricing page (last updated October 22, 2026), including its per-image token counts and the deprecation notice for gemini-2.5-flash-image. Per-1,000 figures and Priority per-image prices are our calculations from those numbers.