Nano Banana 2.1 4K is 3,780 tokens, 3.4x the 1K cost: Google vs Sume

Google prices Nano Banana 2.1 by image tokens: 1,120 at 1K, 1,680 at 2K, 3,780 at 4K. Sume bills a flat tier price instead. Both price ladders, side by side.

5 min readSume
All posts

The short answer

On Google's Gemini API, Nano Banana 2.1 (gemini-nano-banana-2.1) charges $30 per million output image tokens, and a 4K image uses 3,780 tokens against 1,120 for a 1K image. That makes a 4K image 3.375 times the cost of a 1K image: $0.113 against $0.0336 (read 2026-10-09). Sume does not meter tokens for images. It bills a flat price per tier: $0.10 at 1K and $0.20 at 4K, which is 2 times.

Google's token ladder

Google's pricing page gives the token count per output size and says the per-image figures are equivalent prices. A 4096 x 4096 image has 16 times the pixels of a 1024 x 1024 image, but Google charges 3.375 times the tokens, so token cost grows far slower than area.

Nano Banana 2.1 Standard output, Google list vs Sume, as of 2026-10-09
TierGoogle tokensGoogle per imageSume per imageSume tier ratio to 1K
1K (1024x1024)1,120$0.0336$0.101.0x
2K (2048x2048)1,680$0.0504$0.151.5x
4K (4096x4096)3,780$0.113$0.202.0x

Why the two ladders differ

The Google column is Google's own list for its API. The Sume column is what Sume bills to your wallet for the same model id, per image, with no token accounting. The two are priced from different bases, so do not read the Sume column as Google's price plus a fixed percentage. The Sume docs state that usage.cost is the billed USD amount and that token counts in usage are always 0 in v1.

The practical effect is that Sume's 4K is only 2 times its 1K, while Google's is 3.375 times. If you generate mostly at 4K, Sume's tier shape narrows the gap between sizes even though the absolute prices differ.

A worked example

Take a mixed job of 40 images at 1K and 10 images at 4K. On Google Standard the 1K images cost 40 x 0.0336 = $1.344 and the 4K images cost 10 x 0.113 = $1.13, for $2.47 in total. On Sume, the same job costs 40 x 0.10 = $4.00 plus 10 x 0.20 = $2.00, for $6.00. The ratio of the totals is 2.4 times, even though the two ladders have different shapes.

The worked example shows why you should price the size mix you really use. A tier shape that looks flatter does not always give a lower bill, and the larger fixed price per image on Sume buys a single API surface and hosted result URLs.

Choosing between the two ladders

The two ladders answer different questions. Google's per-token rate tells you what one more pixel tier costs in the raw model. Sume's flat tier price tells you what an invoice line will be before you send the request. If you need a hard cap, the flat price is easier to put in a spreadsheet. If you need the lowest list rate and can accept variable token counts, Google's rate is lower on the page.

Whichever you use, test with five images at the tier you plan to ship, read the cost back from the response or the billing console, and only then scale up.

For 100 images, 4K costs $20.00 on Sume (100 x 0.20) and $11.30 on Google Standard (100 x 0.113). At 1K the numbers are $10.00 and $3.36. Those are list prices for two different services, so compare them with your own volume, rate limits, and the way each one hosts and returns files. Sume returns Sume-hosted URLs rather than inline base64, and its 4K calls are among the ones that may return a 202 job instead of a 200.

Sources

Related posts

More in Pricing

All Pricing posts

Written by Sume