Cost per approved banner: 0.5K drafts, 2K final, break-even at 1 in 2

With Nano Banana 2.1 on Sume, drafting at 0.5K ($0.075) then rendering one 2K final ($0.15) beats rendering all at 2K once fewer than 1 in 2 drafts is approved.

4 min readSume
All posts

Drafting at 0.5K ($0.075) and rendering only the approved banner at 2K ($0.15) beats rendering every candidate at 2K as soon as you approve fewer than one in two drafts. At exactly one in two the cost is $0.30 either way; at one in four it is $0.45 against $0.60.

The formula

With an approval rate of 1 in k: draft route = k x $0.075 + $0.15; direct 2K route = k x $0.15. Both are per approved banner, assuming the final is a fresh 2K generation of an approved draft, and that the final matches the draft well enough to be accepted.

Cost per approved banner, Nano Banana 2.1 (read 2026-10-09)
Approval rateDraft at 0.5K, then 2KAll at 2KSaving
1 in 1$0.225$0.15-$0.075
1 in 2$0.30$0.30$0.00
1 in 3$0.375$0.45$0.075
1 in 4$0.45$0.60$0.15
1 in 6$0.60$0.90$0.30
1 in 10$0.90$1.50$0.60

The one number to measure

Your approval rate is the input. Count it across a real batch of 20 drafts before you pick the route: if you keep 12 of 20, render at 2K directly; if you keep 3 of 20, draft first.

def per_approved(k, draft=0.075, final=0.15):
    return k * draft + final

for k in (1, 2, 3, 4, 6, 10):
    print(k, round(per_approved(k), 3), round(k * 0.15, 3))

Gotchas

  • A draft that doesn't look like its final is not a draft. Same prompt, same ratio, new generation: the 2K image can differ.
  • Failed generations are not billed, so retries don't add cost.
  • Use the same aspect_ratio for draft and final; 0.5K to 4K are the tiers Nano Banana 2.1 lists.

Sources

Related posts

More in Pricing

All Pricing posts

Written by Sume