Nano Banana 2.1 image cost on Sume, by resolution tier and count

Nano Banana 2.1 on Sume bills per image at a price set by its resolution tier, 0.5K, 1K, 2K or 4K, so a set of images, from thumbnails to portraits, costs the per-image price times the count. Budget extra for retries when you plan a large set.

Worked example: 100 monster portraits: Nano Banana 2.1 cost on Sume

100 monster portraits cost $10.00 on Sume with Nano Banana 2.1 at 1K (100 x $0.10). The 0.5K tier brings that to $7.50; 2K is $15.00 and 4K is $20.00. With a 20 percent retry allowance the 1K job is $12.00.

A game master prepares portraits for every creature in a campaign and a short intro clip. The open question is how many images to buy and at which size.

What does each tier cost?

Sume bills Nano Banana 2.1 per image by resolution tier. The table prices 100 images at each tier and, for comparison, the same count on Seedream 5 Lite, which Sume lists at one flat per-image price. Both come from Sume's rate card as read on 2026-10-10; re-read the model catalog before a large run, since list prices can change.

How many images should you budget for?

Plan for retries: rejected images are billed like any other. Adding 20 percent to 100 images means buying 120, which at 1K is $12.00. That is $0.12 per kept image if you keep all 100. The 20 percent is an assumption, not a Sume figure; use your own rejection rate.

• Start at 1K ($0.10 per image) and move to 2K only for the images that need the extra detail.
• Use 0.5K ($0.075) for drafts you only need to approve.
• Aspect ratio does not change the Nano Banana 2.1 price in the rate card. If your slot is 3:4, confirm that ratio in GET /v1/images/models before a large run.

What about the old Nano Banana 2 id?

Nano Banana 2 is retired on Sume. Requests that still send google/nano-banana-2 or nano-banana-2 run as Nano Banana 2.1, and the job stores the 2.1 id, so an old script is billed at the 2.1 rates above (Sume image docs, read 2026-10-10).

A request that runs

POST /v1/images takes model, prompt, n (1 to 10), resolution, and aspect_ratio. A request that finishes within 30 seconds answers 200 with image URLs; a slower one answers 202 with a job to poll at GET /v1/jobs/{id}/status. Branch on the status code.

import os
import requests

r = requests.post(
    "https://api.sume.com/v1/images",
    headers={"Authorization": f"Bearer {os.environ['SUME_API_KEY']}"},
    json={
        "model": "google/nano-banana-2.1",
        "prompt": "monster portraits, one consistent style",
        "n": 4,
        "resolution": "1K",
        "aspect_ratio": "3:4",
    },
    timeout=60,
)
print(r.status_code)
body = r.json()
if r.status_code == 200:
    for item in body["data"]:
        print(item["url"])
else:
    print(body)

Where to go next

See every AI model on Sume for the catalog, how Sume pricing works for the billing rules, and the API quickstart for keys and authentication.

Every image set in one table

This page replaces 70 posts that asked the same question with different numbers. Each row is one of them, with the answer it gave when it was published.