Nano Banana API pricing, compared

Price per image — Google list vs APIMODELS, verified 2026-09-26

                                  512px    1K       2K       4K
Nano Banana Pro (Gemini 3 Pro Image)
  Google (list)                   --       $0.134   $0.134   $0.24
  gemini-3-pro-image              --       $0.08    $0.08    $0.13    -40% / -46%
  gemini-3-pro-image-preview      --       $0.068   $0.068   $0.13    -49% / -46%
  gemini-3-pro-image-gemini       --       $0.03    $0.03    $0.03    slow channel

Nano Banana 2 (Gemini 3.1 Flash Image)
  Google (list)                   $0.045   $0.067   $0.101   $0.151
  gemini-3.1-flash-image          $0.04    $0.06    $0.06    $0.10    2K -41%
  gemini-3.1-flash-image-preview  --       $0.05    $0.05    $0.08    2K -50%

Per image, real time. Failed requests are not billed.
Reference (input) images are not charged separately.

cURL

# Step 1: create the task
curl -X POST https://api.apimodels.app/v1/images/generations \
  -H "Authorization: Bearer $APIMODELS_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gemini-3-pro-image",
    "prompt": "A product photo of a ceramic mug on a linen tablecloth, soft window light",
    "aspect_ratio": "16:9",
    "resolution": "2k"
  }'

# Step 2: poll until state is completed
curl "https://api.apimodels.app/v1/images/generations?task_id=TASK_ID" \
  -H "Authorization: Bearer $APIMODELS_API_KEY"

Python

import requests, time

H = {"Authorization": "Bearer YOUR_APIMODELS_KEY"}
URL = "https://api.apimodels.app/v1/images/generations"

task = requests.post(URL, headers=H, json={
    "model": "gemini-3.1-flash-image",   # 2K costs the same as 1K here
    "prompt": "Isometric illustration of a tiny coffee shop, pastel colors",
    "aspect_ratio": "16:9",
    "resolution": "2k",
}).json()["data"]["taskId"]

while True:
    r = requests.get(URL, headers=H, params={"task_id": task}).json()["data"]
    if r["state"] in ("completed", "failed"):
        break
    time.sleep(3)
print(r)

How much does the Nano Banana API cost?

Google lists Nano Banana Pro (Gemini 3 Pro Image) at $0.134 per image at 1K or 2K and $0.24 at 4K, and Nano Banana 2 (Gemini 3.1 Flash Image) at $0.067, $0.101 and $0.151 for 1K, 2K and 4K. On APIMODELS, gemini-3-pro-image is $0.08 / $0.08 / $0.13 and gemini-3.1-flash-image is $0.06 / $0.06 / $0.10, with a lower 1K/2K price on the preview id and a budget option described above.

What is the difference between gemini-3-pro-image and gemini-3-pro-image-preview?

Same Google model family, different channels. The GA id costs $0.08 at 1K or 2K and, over the last 30 days, refused 3.6% of prompts on content grounds. The preview id costs $0.068 but runs on a channel with a stricter filter that refused 12.5%. Refused requests are not billed, but they cost you a retry. Use GA by default and the preview id when price matters more than the occasional refusal.

Why is 2K the same price as 1K on Gemini 3.1 Flash Image?

That is simply how the tier is priced here. Google charges $0.067 for 1K and $0.101 for 2K; here both are $0.06, which makes 2K 41% below Google’s list. If you were rendering at 1K to save money, you can render at 2K for the same price.

How does Google’s Batch API compare?

It is a different product. Batch jobs are queued and can take up to 24 hours to come back, so it only suits offline work that can wait. For anything interactive or user-facing, the real-time prices on this page are the relevant comparison.

What happens to Gemini 2.5 Flash Image?

Google has announced that it will shut Gemini 2.5 Flash Image down on 2026-10-02. It is $0.02 per image here until then, but do not start new work on it. Gemini 3.1 Flash Image is the replacement: better quality, and $0.06 for 1K or 2K.

Do I need a Google Cloud account?

No. Sign up at apimodels.app, create an API key and call https://api.apimodels.app/v1/images/generations. There is no Google Cloud project, billing setup or region selection, it works from mainland China, and the same key and USD balance cover every other image, video and language model on the platform.