Nano Banana 2.1: what's new and what it costs

Nano Banana 2.1 is Google's newest image generation and editing model, released on October 6 and built on Gemini 3.6 Flash. It makes cleaner designs, lets you edit one marked area of a picture, keeps people and products looking the same across edits, and costs about half what Nano Banana 2 did for standard-size images.

Google's launch post on X, from @Google, says the model "outperforms our previous models across the board" and names visual design, mask-based editing and subject consistency as the big jumps. Logan Kilpatrick, who leads Google's AI Studio and Gemini API, sounded more modest. His @OfficialLoganK post calls it "a nice update" with higher quality images, "lots of bug fixes" and a lower price. I think his version is closer to the truth: a solid point release that fixes things and costs less.

What's new in Nano Banana 2.1

Google and the early coverage point to three changes:

• Mask-based editing. You select a region of an image and change only that part. Swapping the background behind a product shot, or fixing one sign in a storefront photo, used to mean regenerating the whole picture and hoping the rest survived.
• Subject consistency. The model accepts up to 14 reference images and, per Google, can keep up to four characters and up to ten objects consistent across a series. That's what you need if the same person, mascot or product shows up in every post of a campaign.
• Design and text. Google says layouts, infographics and text rendering improved, and output now goes up to 4K, including very wide panorama formats.

Decrypt reported Google's internal scores: 1,050 Elo for text-to-image, against 990 for Nano Banana 2. Those are Google's numbers, and nobody had independent benchmarks at launch.

Where you can use it

It's rolling out in the Gemini app, AI Mode in Google Search, Google AI Studio, the Gemini API, Flow, Stitch and Google Ads. In the Gemini app, your plan's existing image limits apply.

Small businesses should keep an eye on the Google Ads placement. Once the model sits inside the tool where ads get built, better and cheaper image variations reach owners who never open an AI app.

What Nano Banana 2.1 costs

Google's Gemini API pricing page lists the model as gemini-nano-banana-2.1, with no free API tier. Per image:

The batch API halves those prices again. At 1K and 2K the cut really is about half. At 4K it's closer to a quarter: Decrypt and The Decoder both printed 7.56 cents for a 4K image, but Google's own pricing page says $0.113. Heise also noted that input tokens and text output now cost more than on the old model, and the pricing page lists $1.50 per million input tokens and $7.50 per million for text and thinking output. If you send long prompts or lots of reference images, look at the bill before you assume it fell by half.

Google's deprecations page names Nano Banana 2.1 as the replacement for gemini-3.1-flash-image, but today it shows no shutdown date for the older model. Some articles cite an October 29 cutoff. Google's docs don't, so I'd treat that date as unconfirmed.

Where it still falls short

According to Heise, Google admits the model still struggles with small text at 1K resolution and with longer passages, and that character consistency is "not always perfect." The Decoder's hands-on tests found that Nano Banana Pro still makes more natural, realistic photos, and that 2.1 sometimes got the relative size of subjects wrong. Every image also carries Google's invisible SynthID watermark.

My read: for social graphics and ad variations, 2.1 is now the obvious pick on price. If the job depends on photorealism, say a hero image of a finished kitchen, keep testing it against Pro.

What this means for a small business

At 2K, a hundred social images cost about five dollars through the API. The money stops mattering at that point, and the work around the image becomes the cost: someone has to write the prompts and someone has to check the text on each image before it goes live. A trade shop posting job photos or a salon running weekly promos can get real use out of this, as long as that routine gets set up once instead of reinvented every week.

If you're not sure where image tools would save time in your business, New Face Design's free process audit starts with what you post and sell today, before anyone picks a model.