$200 of image credit at 2K: 1,333 Nano Banana 2.1 or 3,595 GPT images
A $200 grant buys 1,333 Nano Banana 2.1 images at 2K or 3,595 GPT Image 2.5 medium images on Sume. Google lists 3,968 standard and 7,936 batch.

The short answer
A $200 budget buys 1,333 Nano Banana 2.1 images at 2K on Sume ($200 divided by $0.15 = 1,333.3, rounded down) or 3,595 GPT Image 2.5 medium images at 2K ($200 divided by $0.055625 = 3,595.5, rounded down). Google's page implies 3,968 at its standard 2K rate and 7,936 at its batch rate.
Nonprofits and grant-funded projects often have a fixed amount, so counting images is more useful than quoting a price.
The arithmetic
Each count is the budget divided by the unit price and rounded down, because Sume bills in whole images and a partial image is not sold. The table uses 2K for all rows so that the comparison is like for like.
At 1K, the Sume Nano Banana 2.1 count is 2,000 ($200 divided by $0.10), and at 0.5K it is 2,666.
| Option | Arithmetic | Images |
|---|---|---|
| Nano Banana 2.1 at 2K (Sume) | $200 / $0.15 | 1,333 images |
| GPT Image 2.5 medium, 2K (Sume) | $200 / $0.055625 | 3,595 images |
| Nano Banana 2.1 at 1K (Sume) | $200 / $0.10 | 2,000 images |
| Nano Banana 2.1 at 0.5K (Sume) | $200 / $0.075 | 2,666 images |
| Google direct, 2K standard (reference) | $200 / $0.0504 | 3,968 images |
| Google direct, 2K batch (reference) | $200 / $0.0252 | 7,936 images |
How to run it
Plan a reserve of 10 percent for retries, which on the Nano Banana 2.1 2K row is about 133 images. The failed generations on Sume are not billed, but the ones you reject by eye are.
Here is the same arithmetic in code, so that you can change the budget and the price and see the count.
from decimal import Decimal as D
budget = D("200")
for name, price in [
("nano-banana-2.1 2K", D("0.15")),
("gpt-image-2.5 medium 2K", D("0.055625")),
("google 2K standard", D("0.0504")),
("google 2K batch", D("0.0252")),
]:
print(name, int(budget // price))Retries and what to watch
A count is only as good as the retry rate you assume. At a 25 percent reject rate, the usable counts are 80 percent of the figures in the table, because four images are bought for every three you keep.
Applying that to the Sume rows gives the usable counts below.
- Nano Banana 2.1 at 2K: 1,333 images bought, about 1,066 kept at 80 percent.
- GPT Image 2.5 medium at 2K: 3,595 images bought, about 2,876 kept at 80 percent.
- Google 2K standard (reference): 3,968 bought, about 3,174 kept at 80 percent.
What the docs say to check
The endpoint pricing lines are what Sume charges your wallet and already include the Sume margin, so a call costs cost_usd times n. Sume bills image generation all or nothing: a generation that fails or is cancelled is not charged.
The counts change if the price table changes. Re-read the public catalog on the day you commit the money.
The numbers above are list prices read on 2026-10-08, not measurements. Sume figures are the catalog prices that already include its margin, Google figures are from its own pricing page, and quality or speed differences are not covered. Re-read both pages before you commit a large budget, because either side can change its rates.
Sources
Related posts
More in Pricing
- 200 support calls of 6 minutes: 20 hours of transcripts, priced
20 hours of support calls cost $10.80 at MAI-Transcribe-2-Streaming's intro price and $12.00 on Sume STT at $0.01 a minute. Job count and limits included.
- 23-second hero shot: the 720p to 1080p step costs $2.88 or $19.41
For a 23 s hero shot, going from 720p to 1080p adds $2.875 on Wan 3.0 ($2.875 to $5.75) and $19.41 on Seedance 2.5 ($13.29 to $32.70).
- 24 field interviews of 25 minutes: 10 hours transcribed
24 field interviews of 25 minutes are 10 hours: $5.40 on MAI-Transcribe-2-Streaming's intro price and $6.00 on Sume STT in 72 jobs of 10 minutes or less.
- 24 product demo recordings of 8 minutes: transcript cost
24 demo recordings of 8 minutes are 3.2 hours: $1.73 at MAI-Transcribe-2-Streaming's intro price and $1.92 on Sume STT, one job each.
Written by Sume