Nano Banana 2 Lite has no Search grounding; NB2 does

If a prompt needs live facts, such as today's scoreboard or a new product's look, use Nano Banana 2, not Lite. Google's image-generation docs say NB2 supports grounding with Google Search and image search, and Lite does not. Sume does not expose a grounding option and does not list a Lite row, so on Sume the closest catalog row is google/nano-banana-2, whose static list rate is $0.08 per image. Nano Banana 2 Lite is not in the Sume image catalog today.

Which Gemini image id does what

Facts below are from Google's image-generation page.

When it matters

Grounding lets the model look things up before it draws. Without it, the model draws from training data only, which is fine for a generic product shot and riskier for anything tied to recent events.

• Needs current facts: NB2 on the Google API.
• Static prompt with no real-world lookups: Lite is described by Google as the fastest and most cost-efficient in the family.
• Not sure: run the same prompt on both and compare.

On Sume

The Sume Image API schema has no grounding parameter. A request that sets a parameter the model does not list returns 400 unsupported_parameter. So on Sume, put the facts in the prompt yourself. Verify the model with GET /v1/images/models, and quote the billed price from the pricing array: $0.08 list x 1.25 = $0.10 for google/nano-banana-2.