Nano Banana 2.1 edit vs generate on Sume: same price per image by tier

On Sume, a Nano Banana 2.1 edit with reference images costs the same as a text-to-image call at the same resolution tier: $0.075 at 0.5K, $0.10 at 1K, $0.15 at 2K and $0.20 at 4K per image. The cost estimator for this row reads the tier and the image count (n), not the number of references.

The table

Where edits can cost more

Not every row is flat. The docs describe ChatGPT Image 2.5 at Fal token rates, with $8 per million input image tokens and $5 per million input text tokens, so images you send in add to the bill. Nano Banana 2.1 is one price per output image. If you compare models for an edit-heavy job, read each model's pricing lines from GET /v1/images/models/{model_id}/endpoints; those lines already include the Sume margin.

curl "https://api.sume.com/v1/images/models/google/nano-banana-2.1/endpoints" \
  -H "Authorization: Bearer $SUME_API_KEY"

Gotchas

• aspect_ratio: "auto" matches the reference; leaving the field out is not the same.
• A failed edit is not billed.
• The billed amount is usage.cost on the response; reconcile against price x n.