Nano Banana 2.1 4K is 3.4x the 1K price at Google, 2x on Sume

On Google's own pricing page, a Nano Banana 2.1 4K image costs 3.375 times a 1K image: 3,780 output tokens against 1,120. On Sume the same two tiers are $0.20 and $0.10, a 2x gap. The gap exists because Google charges per output token and Sume lists a per-image price per tier.

Google's token math

Google's page says image output is $30 per 1,000,000 tokens on the standard tier. A 1K (1024x1024) image uses 1,120 tokens, 2K uses 1,680 and 4K uses 3,780. That gives $0.0336, $0.0504 and $0.113 (3,780 x 30 / 1,000,000 = $0.1134).

Sume's tier prices

Sume's pricing tables carry provider prices of $0.08 at 1K, $0.12 at 2K and $0.16 at 4K for google/nano-banana-2.1, then bill them times 1.25. Those are different provider prices from Google's direct page, so Sume's ratios are 1.0, 1.5 and 2.0.

What that means for choosing 4K

On Sume, 4K is the better buy per pixel: a 4096x4096 frame has four times the pixels of 2048x2048 and costs $0.05 more than 2K. If you need a print or a large hero, go straight to 4K rather than rendering 1K twice. Remember that the docs say 4K and large n are the most likely to return a 202 job envelope instead of an inline image, so handle both codes.

Compare before you build

Do not infer Sume's price from Google's. Read the pricing array from /v1/images/models/google/nano-banana-2.1/endpoints and bill from what it returns. See the Google vs Sume gap for the full comparison.

Worked example: 200 hero images

Take 200 hero images. At Google's per-token prices a 4K batch is 200 x $0.113 = $22.60 (using the rounded per-image figure), while a 1K batch is 200 x $0.0336 = $6.72. On Sume the same two batches are 200 x $0.20 = $40.00 and 200 x $0.10 = $20.00. The difference between 4K and 1K is $15.88 at Google and $20.00 on Sume.

That shows why you should compute from the price you will actually pay. If you plan the budget from Google's ratio of 3.375 and bill on Sume, you overestimate the cost of 4K relative to 1K. If you plan from Sume's 2x and call Google directly, you underestimate it.

Also note what Google lists for batch: its page prices the batch tier at half the standard rate, $0.0168 for 1K, $0.0252 for 2K and $0.0567 for 4K. Sume's catalog lists one per-image price per tier and no batch tier, so the batch discount is a Google-direct feature.

Google's page lists Nano Banana 2.1 at 1K, 2K and 4K, with a standard rate and a batch rate, and says image output is billed in tokens. It also lists search grounding as a separate charge: 5,000 free requests a month shared across Gemini 3.x models, then $14 per 1,000. None of that is in Sume's per-image price.

Sume's parameter list for the row has no grounding field, so a Sume image call never pays for search queries. If you rely on grounded images, call Google directly. If you rely on a flat per-image price and one HTTP API for many models, the Sume tiers are easier to budget.