Nano Banana 2.1: Google vs Sume price for one 1K image

For one 1K Nano Banana 2.1 image, Google's Gemini API pricing page lists $0.0336 on the standard tier, while Sume bills about $0.10, which is the Sume catalog list price of $0.08 times 1.25. Your Sume invoice follows the Sume number, and it is the one that appears in usage.cost.

The gap looks large, so it is worth understanding where it comes from before you pick a place to buy from.

The numbers

Why they differ

Google bills by tokens, so the price can move with prompt length and settings. Sume prices per image with a fixed catalog rate by resolution tier, and the repository pricing table carries $0.08 at 1K, with the margin of 1.25 on top. Sume bills from its own table, so a vendor price cut does not automatically reach your bill. Read the pricing array from GET /v1/images/models/google/nano-banana-2.1/endpoints for the figure that applies.

What you buy on Sume is a different package: one API shared with the other rows, per-image prices you can budget without token math, hosted result URLs, and a failed generation that is not billed.

When the cheaper source wins

• A steady, very large batch on one model with your own Google account: Google's batch tier at $0.0168 per 1K image is far below Sume.
• Mixed models in one pipeline, with an agent choosing: one router and one bill is simpler than several vendor accounts.
• A pilot or a campaign of a few hundred images: the dollar difference is small next to engineering time.

Budget it both ways

Take 500 images at 1K. At Google's standard rate that is $16.80; at Sume's billed rate it is $50.00. If you need 500 images, the $33 gap pays for a lot of integration time; if you need 500,000, it does not. Do this arithmetic with your own volume before you choose, and use the cheaper source when volume is high and your stack is single-vendor.

Remember that Google's price depends on prompt and thinking tokens, so test a real prompt. Sume does not expose the thinking level, so you cannot lower it to save money on this route.

Check your own invoice

Send one request at each tier you use and note usage.cost from the response. Multiply by your monthly count and compare with the vendor page's figure for the same count. If the gap is under what a week of integration work costs, stay on one API; if it is far above, consider a direct account for that one workload and keep Sume for the rest.

Revisit the comparison when either side changes its prices, since both pages are dated 2026-10-08 in this post.

What Sume does not do

Sume does not match vendor list prices or offer a batch discount on this row. It also does not show the upstream cost to you; only the billed amount. The Sume figure in this post comes from the catalog on 2026-10-08, so re-read the endpoints route before an important budget.