Google released Nano Banana 2.1 on Tuesday, wiring the new image model into Google Ads, Google Search’s AI Mode, the Gemini app, Google AI Studio, Flow and Stitch on day one, with a 1K image now priced at $0.0336 on the Gemini developer API, down from $0.067 on Nano Banana 2. A 4K image falls to $0.0756, from $0.151. Batch jobs carry an additional 50 percent discount, pulling a 1K image to $0.0168. The higher-end Nano Banana Pro still runs $0.134 per 1K image, and Google frames the new release as its “more efficient counterpart.”
Paid-tier input runs $1.50 per million tokens across text, image and video; text and thinking output is $7.50 per million, image output $30.00. Grounding with Google web and image search is free for the first 5,000 requests per month across Gemini 3.x models, then $14 per 1,000. The Gemini 3.1 Flash Image API shuts down as early as October 29.
That’s the surface. The structural story is that Google is compressing the production cost of advertising creative inside the ad surface where small businesses already spend money.
Nano Banana 2.1, based on Gemini 3.6 Flash and shipped as gemini-nano-banana-2.1, posts a 1,050 ±14 text-to-image preference Elo against 990 ±7 for its predecessor and 935 ±8 for Nano Banana Pro, per the DeepMind model card. Multi-character consistency climbs to 1,106 from 978. Infographic factuality moves to 0.521 from 0.179. Product consistency sits at 1,024, mask-based editing at 1,049. The model accepts up to 14 reference images, tracks 4 characters and 10 objects, and outputs at 1K, 2K or 4K in aspect ratios as wide as 21:9. RuntimeWire notes the scores are Google’s internal results, not an independent benchmark, and flags that the API docs call the model an update to Gemini 3.1 Flash Image while the model card cites Gemini 3.6 Flash.
The original Nano Banana, released September 2025, pushed Gemini to the top of both major app stores on the back of selfie-to-figurine novelty, per Decrypt. The sequel is a quieter business story. For a small-team operator running Google Ads, $33.60 now buys a thousand 1K images where $67 did before, inside the same ad console. The creative-cost barrier that kept visual campaigns out of reach is largely gone.
What the new pricing doesn’t change: which products to feature, which audiences signal present buying intent, what offer converts. The upstream decisions still require owner hours or a dedicated marketing hire. Cheaper pixels, same strategy gap. Services like LemonLime, which prepare proactive customer-growth work rather than wait to be prompted, sit squarely in that gap, alongside Google’s recent moves to open its AI shopping checkout hub to all U.S. merchants and to lock small businesses into AI Max.
The creative layer is a commodity now. The judgment layer isn’t.
Sources
- https://deepmind.google/models/model-cards/nano-banana-2-1/
- https://the-decoder.com/googles-new-image-model-nano-banana-2-1-generates-better-images-for-less-money/
- https://www.unite.ai/nano-banana-2-1-debuts-at-half-the-image-cost-of-its-predecessor/
- https://decrypt.co/380257/google-launches-nano-banana-2-1
- https://runtimewire.com/article/google-nano-banana-2-1-image-model
