Nano Banana 2.1 1K image is 1,120 tokens: Google's $0.0336 explained

Google's pricing page says Nano Banana 2.1 output images at 1K consume 1,120 tokens and image output costs $30 per million tokens on Standard, so a 1K image is 1,120 x $0.00003 = $0.0336. Batch is exactly half at $0.0168. The 2K and 4K prices imply 1,680 and about 3,767 tokens.

The arithmetic

The implied token counts divide each listed price by $0.00003. Only the 1K count (1,120) is stated on the page, so the other two are derived, and the 4K figure is rounded.

Why this helps

Knowing the token basis tells you how the price scales: 2K is 1.5 times the 1K price (0.0504 / 0.0336), and 4K is about 3.36 times. If Google changes the per-token rate, all three rows move together, so one number to watch is $30 per million.

It also explains why Sume does not mirror these rows. Sume bills per image from its own catalog: $0.075 at 0.5K, $0.10 at 1K, $0.15 at 2K and $0.20 at 4K. The 4K to 1K ratio on Sume is 2.0, not 3.36, so the two price lists are not proportional.

Using Sume's per-image rows

Sume's docs say token counts in image responses are always 0 in v1, because image models are metered per image, and usage.cost is the USD amount billed. So on Sume you budget with the row, not with tokens: 100 images at 2K are 100 x $0.15 = $15.00.

For the same 100 at Google Standard, 100 x $0.0504 = $5.04, or $2.52 on Batch. That is the price of calling Google directly. Choose Sume when one wallet and one request shape across models matter more than the lower direct rate.

Check your own bill

If you call Google directly, multiply images by the tier price from this table and compare with your invoice. A mismatch usually means the images were generated at a different tier than you assumed or that part of the traffic went through Batch.

On Sume, compare usage.cost on the response with the row for the model and tier. They should be equal for a single image, and the sum over a run should equal images times the row.