Nano Banana quality parameter: why quality returns 400

Nano Banana 2 and Nano Banana Pro on Sume have no quality parameter. Sending one returns 400 unsupported_parameter, because Sume rejects any parameter the chosen model does not advertise instead of dropping it silently. To change how much detail you get, change resolution (512, 1K, 2K, 4K) or move between the two models.

This is from the Image API docs and the catalog code, read 2026-09-29.

Why does Sume reject quality on Nano Banana?

The image catalog advertises quality only for models that have quality settings. A model with none advertises nothing, and a request that names it is refused. The docs put it plainly: parameters a model does not list return 400 unsupported_parameter rather than being silently dropped.

Which Sume image models take quality?

What does the 400 look like?

In current code the error body carries field: "quality" and a catalog_url of /v1/images/models.

curl -X POST https://api.sume.com/v1/images \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "google/nano-banana-pro",
    "prompt": "A glass perfume bottle on wet slate",
    "quality": "high"
  }'

How do I get more detail from Nano Banana instead?

Raise the resolution tier and re-run. Both models list the tiers above, and the list price scales with the tier in the catalog code, so check the estimate before you send a 4K batch. Nano Banana Pro and Nano Banana 2 are separate models; if the look is the issue, try the other one rather than a parameter.

curl -X POST https://api.sume.com/v1/images \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "google/nano-banana-pro",
    "prompt": "A glass perfume bottle on wet slate",
    "resolution": "2K",
    "aspect_ratio": "4:5"
  }'

Does quality auto work on Nano Banana?

No. auto is a value of quality, and Nano Banana has no quality at all, so auto is rejected the same way as high. The auto value is listed for GPT Image 2.5 only. If a client library sets quality by default, override it per model, or omit the field entirely for the Nano Banana ids.

For GPT Image 2.5 the docs say omitted quality defaults to high, so a client that expects that default on Nano Banana will not get it: the field simply does not exist there. Leaving quality out does not trigger this error; keep the other parameters to the ones the catalog lists for the model.

How do I avoid the error in a shared client?

If one client sends the same body to several models, read GET /v1/images/models once and send only the parameters each model lists. That keeps a quality default meant for GPT Image from reaching a Nano Banana call. Log the model id next to the body you sent, so a 400 points at the exact pairing that caused it.