Nano Banana 2.1 thinking levels and search grounding: not on Sume

No: Sume's google/nano-banana-2.1 row does not expose thinking levels, web-search grounding or a system prompt. fal's model page documents all three: thinking levels of minimal, medium and high, optional web search grounding, and a system prompt for style control. On Sume, a field the model's catalog entry does not list is rejected with 400 unsupported_parameter, not silently dropped.

fal's page vs the Sume row

The Sume side comes from the catalog capabilities in the repo, which generate the supported_parameters descriptors. They cover prompt, aspect ratio, resolution, n, output format and references.

What to do instead

Put the style rules in the prompt text, because the prompt is the one field that always passes. For facts that must be current, fetch them first and paste them into the prompt; the model will not search for you on Sume.

Confirm what a row accepts with one request to its endpoints record:

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

A reminder on ids

The old google/nano-banana-2 id still works and runs as 2.1, and the job stores the 2.1 id. The base list rate is $0.08 per image at 1K, so $0.10 billed; larger tiers scale on admission. Read the rate card for current credits.