Nano Banana 2.1: Google Halves Image Prices, Lifts Quality

Nano Banana 2.1: Google Halves Image Prices, Lifts Quality

Google has shipped Nano Banana 2.1, its newest model for generating and editing images. The headline change is the price: roughly half of what the previous version cost. Google also says the model is better than its predecessors "across the board."

It replaces Nano Banana 2, which arrived in February 2026. Developers still using the older model should note the deadline: Google will switch off "gemini-3.1-flash-image" on October 29, 2026.

What sits under the hood

Nano Banana 2.1 is built on Gemini 3.6 Flash, the fast, lightweight tier of Google's Gemini family. That makes it a different product from Nano Banana Pro, which runs on the newest Pro model, version 3.1. Pro is probably still Google's strongest image model.

Google calls 2.1 "the more efficient counterpart" to Pro. In practice, the company is offering two lanes: a fast, cheap model for most work and a more expensive one for the best possible output.

That split may not last in its current form. Google has already announced its next Pro model, Gemini 4 Argon, and the next big step in image quality could come from there.

The new price list

The cuts are substantial. Here is how the per-image costs compare:

  • 1K image, Nano Banana 2.1: 3.36 cents (down from 6.70 cents)
  • 4K image, Nano Banana 2.1: 7.56 cents (down from 15.10 cents)
  • 1K image, Nano Banana Pro: 13.40 cents

A single 1K image from Pro now costs about four times as much as one from 2.1. Even a 4K image from 2.1 is cheaper than a 1K image from Pro.

What Google says is better

According to Google, the update brings improvements in several areas:

  1. Overall visual quality. Images should look better in general.
  2. Text rendering. Lettering inside images, long a weak spot for image models, is said to be cleaner.
  3. Character consistency. People and figures should look the same from one conversation turn to the next.
  4. Wide panoramas. Broader scenes are handled better.
  5. Infographics. Charts and explanatory graphics are on the list too.

The model accepts up to 14 reference images in one go. From those, it can keep up to four characters and ten objects consistent across outputs. That matters for anyone building a series of images, such as a storyboard or a set of product shots, where the same person or item must appear repeatedly.

Nano Banana 2.1 can also pull from Google Search. And it offers three thinking levels - minimal, medium and high - which affect the quality of the final image. More thinking generally means better results, with the trade-off you would expect in speed.

Supported resolutions are 1K, 2K and 4K, and the model runs at Flash-tier speed.

Benchmarks versus real use

On paper, 2.1 sometimes beats Pro by a wide margin. But there is a catch: Nano Banana 2 also matched Pro in those same tests. In everyday use, Pro often still delivers visibly better images.

That gap between benchmark scores and what users actually see is worth keeping in mind before treating 2.1 as a Pro replacement. It is cheaper and faster. Whether it is good enough depends on the job.

Where you will find it

Google is rolling the model out across a long list of its products:

  • the Gemini app
  • AI Mode in Google Search
  • Google AI Studio
  • Flow by Google
  • Stitch by Google
  • Google Ads
  • the Gemini Enterprise Platform

Further technical details are in the model card Google has published.

Our Take

The most important number here is not a benchmark score. It is the price cut. Halving the cost of an image changes the math for anyone generating visuals at scale, from ad teams inside Google Ads to developers building on AI Studio. This suggests Google sees volume, not just peak quality, as the way to win the image market.

The two-tier setup also looks deliberate. Flash for everyday work, Pro for the hardest cases. Google has drawn a similar line on the chatbot side, where it recently limited free users to Flash-Lite. Cheaper models handle the bulk; premium models are reserved for those who pay.

Competition is pushing in the same direction. Rivals such as Flux 3 Image are also focusing on consistency across multi-step edits, the same problem Google is targeting with its reference-image support.

It is worth watching whether independent testers find 2.1 closing the real-world gap with Pro, or whether the benchmark gains once again fail to show up in practice. The other open question is what Gemini 4 Argon brings to image quality when it arrives.