Nano Banana 2 Lite alternative on Sume: image rows under 4 cents

Sume's image catalog has no Nano Banana 2 Lite row. The Nano Banana rows it does list are Nano Banana 2, which prices from $0.0750 at its 0.5K tier, and Nano Banana Pro at $0.1875 up to 2K. If you were looking for a cheaper Nano Banana, the closest thing on Sume is a different model family, and seven rows quote under four cents per image.

The table below lists them with the features that decide whether a row can replace a Nano Banana call. Prices are Sume's, which are provider list prices times 1.25, and the GPT Image 2.5 rows are quotes at 1024x1024 with a short prompt.

Rows under $0.04

Sorted by price. The fourth column marks edit support, since many Nano Banana jobs are edits with a reference. A flat price means the figure does not change with the size or any option.

Which row fits which job

If the job is an edit with one or two references and you want the lowest price, GPT Image 2.5 at low or medium is the cheapest edit row: a one-reference edit at 1024x1024 quotes $0.0094 at low and $0.0209 at medium. Grok Imagine, Qwen Image, Seedream 4.0 and Flux 2 Pro take references too and keep a flat price whatever the size.

If the job is text-to-image only, Imagen 4 Fast is the same price as Grok Imagine and Qwen Image and does not do edits, so it only makes sense when you want that model's look. Seedream 4.0 and Flux 2 Pro are the two rows closest to four cents and the nearest in price to the higher GPT Image 2.5 tiers.

What you give up against Nano Banana 2

Check these before you move a workload.

• Aspect ratios. Nano Banana 2 lists 15 ratios in Sume's catalog. Grok Imagine and Qwen Image list 13, and Imagen 4 Fast lists 5.
• Resolution tiers. Nano Banana 2 offers 0.5K, 1K, 2K and 4K as named tiers. The flat-price rows above do not list resolution tiers, so size is set by the model.
• Output count. Every row above except Grok Imagine can return up to four images per call, the same as Nano Banana 2.
• Look. The price says nothing about whether the image matches what you get from Nano Banana 2, so run your own prompts.

Find the live list

Do not copy this table into code. Ask the catalog, which returns every model id and its price, then filter.

curl -s https://api.sume.com/v1/images/models \
  -H "Authorization: Bearer $SUME_API_KEY" \
  | python3 -c "
import json, sys
body = json.load(sys.stdin)
rows = body.get('data', body) if isinstance(body, dict) else body
for row in rows:
    print(row.get('id'))
"

A note on Lite names

Model names change faster than catalogs, and a Lite variant could appear on a provider's own site before it appears in Sume's. When a name you expect is missing, trust the live response from GET /v1/images/models over any list of names, including this one, and check usage.cost after your first call.

Sume accepts the bare Image Router ids, such as nano-banana-2, as aliases for their org/slug form, so a request that names an id the catalog does not list returns an error, not a silent substitute.

If you need the Nano Banana look at a lower price, there is no row on Sume that offers it. The honest options are to accept a different model at under four cents, or to stay on Nano Banana 2 at 0.5K, where one image costs $0.0750. At 1024x1024 a GPT Image 2.5 medium image is 4.5 times cheaper than that smallest Nano Banana tier, so the saving from switching is large enough to justify a proper blind test of the output.