Nano Banana 2.1 costs $0.0336 at Google and $0.10 on Sume: the gap

Google lists Nano Banana 2.1 at $0.0336 per 1K image on its Gemini API pricing page, and Sume's card is $0.10. For image generation alone, calling Google directly is about three times cheaper at 1K.

At 4K the gap narrows to 1.8 times ($0.113 against $0.20). Sume's price makes sense when you chain video, captions and Timeline through one job API, not when you only want pictures.

Side by side

Google's numbers are from the Gemini API pricing page, model id gemini-nano-banana-2.1. Sume's are the card on the API pricing page.

Why the numbers differ

Sume bills provider list times 1.25. Dividing $0.10 by 1.25 gives $0.08, which is not Google's $0.0336, so the upstream rate Sume pays is not Google's direct Gemini API rate. Sume's docs do not break that down further, and I will not guess at it.

Treat the Sume number as a platform price: you get an asynchronous job, a stored result URL and one catalog for images, video and audio. You do not get Google's rate.

A volume check

At 10,000 images a month at 1K the direct rate is $336 and the Sume card is $1000, a gap of $664 a month. That is the number to weigh against the engineering time of running your own queue, storage and retries.

Below a few hundred images a month the gap is a few dollars and the single job API usually matters more. Above that, the gap deserves a real comparison.

When to go direct

If your workload is a high volume of Nano Banana 2.1 stills, such as 1,000 product images at 1K, Google direct is $33.60 and Sume is $100.00. Take the direct route if you already have a Gemini API account and a place to store files.

Sume is the better fit when the same job needs a clip from the image, captions and a Timeline assembly, and you want one key, one billing line and one job pattern. The 0.5K size ($0.075) has no Google rate to compare with on the page I read.

Prices are as read on 2026-10-07. Both sides can change; check the live pages.