Nano Banana 2 Lite vs Nano Banana 2 per 1K image: $0.0336 vs $0.067
Google lists a 1K Nano Banana 2 Lite image at $0.0336 and Nano Banana 2 at $0.067. What Lite drops, and which Banana ids Sume lists.

Google's Gemini API prices a 1K image from Nano Banana 2 Lite (gemini-3.1-flash-lite-image) at $0.0336 and one from Nano Banana 2 (gemini-3.1-flash-image) at $0.067, so Lite is about half the price, at 1K only. At 1,000 images that is $33.60 against $67. The saving comes with limits: Lite has no 512px, 2K or 4K tier, and no Google Search grounding.
Prices are from the Gemini API pricing page and capabilities from the Nano Banana image generation page, both read 2026-10-02.
What are Google's per-image rates?
The pricing page lists Nano Banana 2 at $0.045 for a 0.5K image, $0.067 for 1K, $0.101 for 2K and $0.151 for 4K. Lite is $0.0336 for 1K. The token prices behind them are $60 per 1M output image tokens for Nano Banana 2 and $30 for Lite, which is why the ratio is close to one half.
| Model | Tier | Per image | 1,000 images |
|---|---|---|---|
| Nano Banana 2 Lite | 1K | $0.0336 | $33.60 |
| Nano Banana 2 | 0.5K | $0.045 | $45.00 |
| Nano Banana 2 | 1K | $0.067 | $67.00 |
| Nano Banana 2 | 2K | $0.101 | $101.00 |
| Nano Banana 2 | 4K | $0.151 | $151.00 |
What do you give up with Lite?
Google's image page lists Lite as 1K only, with no Google Search grounding, and with reference limits of 14 objects and no character or style slots. Nano Banana 2 adds 0.5K, 2K and 4K, Google Search and image search grounding, and character and style references. Thinking levels (minimal, the default, and high) apply to both Flash and Lite per the page.
So Lite fits high-volume 1K work where the prompt carries everything: thumbnails, catalogue variants and drafts. It is the wrong choice for a 4K hero image or a prompt that needs fresh facts from search.
What does Sume list?
Sume's image catalog is read from GET /v1/images/models, and each row reports its own resolution tiers, aspect_ratio list and input_references range. In the docs I read, Sume's resolution tiers are 512, 1K, 2K and 4K, and Banana models appear as nano-banana-2 and nano-banana-pro; I did not find a Nano Banana 2 Lite id. If you specifically want Lite pricing, that is a Google API route, not a Sume one today.
Sume's billing is per image from the endpoint's pricing line with margin applied, so cost_usd times n is the charge, as the Images docs state. Our Nano Banana 2 pricing post lists Sume's tier prices.
How do you check your own cost?
Request the endpoint record and read the pricing array for the id you plan to call, then multiply by your monthly image count. Do this at the resolution you will use, since 4K costs far more than 1K on Google's page.
curl "https://api.sume.com/v1/images/models/nano-banana-2/endpoints" \
-H "Authorization: Bearer $SUME_API_KEY"When does the cheaper image cost more in the end?
Price per image ignores retries. If Lite needs two attempts to hit a result that Nano Banana 2 lands in one, you pay $0.0672, which is more than one Nano Banana 2 image at $0.067. Measure your own keep rate on 50 prompts before choosing on price alone.
The same logic applies to grounding. Lite has no Google Search grounding, so any prompt that depends on current facts needs those facts pasted into the prompt, and that costs your time.
What is the takeaway?
If 1K is enough and you call Google directly, Lite halves the bill. If you need larger tiers, grounded prompts or character references, Nano Banana 2 is the one to price. Re-check both rates before a large run; Google's pricing page is the source of truth, not this table.
Sources
Related posts
More in Pricing
- organization_wallet_not_provisioned vs insufficient_credits (402)
A Format run can fail at create with 402 for two reasons: an empty wallet or an unfunded organization wallet. How to tell them apart and who fixes each.
- generation_admission_preview: check before paid hosted MCP calls
Sume's hosted MCP has generation_admission_preview, dry_run and max_spend_usd. What each one checks, when to use it, and when a plain single create is fine.
- Can a per-run spend cap raise the limit? Formats yes, schedules no
Formats honor a per-run generation_spend_cap_usd above the Format cap, schedules clamp it, and Agent Completions require it. Defaults and edge cases.
- Rask AI minute credits and 3x lip-sync vs a Sume dubbing pipeline
Rask bills 1 credit per video minute for Standard lip-sync, 3 for Enhanced. Sume has no dubbing endpoint; a chained pipeline runs about $0.55-$1.55 per 10 min.
Written by Sume