Nano Banana 2.1 in Google Flow

Nano banana 2.1 in Google Flow: it is official now

The Flow story ran ahead of the launch by about two weeks. A Google Flow web build referenced a model called "Nano Banana 2.5 Flash" on 2026-09-24. On 2026-09-27 the same string was relabeled "Nano Banana 2.1". On 2026-10-05 it became a selectable option in Flow's image model picker. On 2026-10-06 Google released it properly and made it generally available.

So what testers were poking at in early October was the real thing, not a routing experiment. The relabel theory that went round at the time, that 2.1 might be Nano Banana Pro under a new name because the outputs looked similar, turned out to be wrong: 2.1 is a new Flash-line model, an update to Nano Banana 2, and Pro still exists alongside it.

What that means in Flow is simple. The picker entry now has a model behind it that Google documents, prices and supports, so anything you make with it is production work rather than a tester's curiosity.

How to pick Nano Banana 2.1 in Flow

Flow organizes work into projects, and the model is a per-step choice rather than an account setting. The flow of it:

• Open Google Flow and sign in with the Google account you want the work saved under.
• Open an existing project or start a new one. Everything else happens inside a project.
• Find the image model picker for the step you are working on, and choose Nano Banana 2.1 from the list. It sits with the other image models, including Nano Banana Pro and Nano Banana 2 Lite.
• Set the controls Flow exposes for that step, such as aspect ratio and output size, then write your prompt and generate.
• Switch models between attempts if you want a comparison. The picker is per step, so one project can hold a Flash draft and a Pro final.

Flow's interface moves, and Google changes what each step exposes, so treat the labels above as a shape rather than a script. The part that does not change is where the choice lives: in the model picker attached to the step, not in a settings page.

Every output carries a SynthID watermark, as with Google's other image models. That is not a Flow setting and there is nothing to turn off.

What Google Flow is, in plain terms

Google Flow is Google's creation tool for video and images. You work inside a project, write prompts, and choose which model handles each step. It is built for people making things rather than for developers, so it gives you a UI and a fixed set of controls instead of parameters and an API key.

Two consequences are worth knowing. First, model names in Flow are product labels, and product labels do not have to match API model ids. That is exactly why this family has nicknames: the friendly names and the Gemini names diverged, and "Nano Banana 2.1" in Flow is gemini-nano-banana-2.1 in the API. Second, availability in Flow rolls out gradually, so two people on the same day can see different pickers.

Flow is also where new image models tend to appear first, which is why it gets watched so closely. The 2.1 sequence is a clean example: in the build, then in the picker, then announced.

Reference sheet

Four-frame sequence

Nano Banana 2.1 in AI Mode: the banana icon

The other consumer route is AI Mode in Google Search, inside the Google app. The entry point is a banana icon under the search box: tap it, describe the image you want or attach one to edit, and the generation runs there.

This is the quickest way to try the model if you do not want a project, a canvas or an API key. It is also the most limited: you get the controls that AI Mode chooses to show, not the full parameter set that the API exposes, so resolution, aspect ratio and thinking level are not necessarily yours to pick.

Availability here depends on your account and your country, and Google has not published either the country list or the per-plan limits. If the banana icon is not under your search box, that is a rollout answer, not a setting you missed.

Where else Nano Banana 2.1 runs

Google shipped 2.1 across its products on release day, so Flow is one option among several:

• The Gemini app, for ordinary prompting and image editing.
• AI Mode in Google Search, through the Google app, behind the banana icon.
• Google AI Studio, which is the easiest place to try prompts and parameters against the real model without writing code.
• Google Flow, for project-based image and video work.
• Stitch, for interface and product design.
• Google Ads, for ad creative.
• Gemini Enterprise, for business accounts.
• The Gemini API, as gemini-nano-banana-2.1, priced at $0.0336 per image at 1K, $0.0504 at 2K and $0.0756 at 4K, with batch mode 50 percent off.

If what you want is to see what the model can actually do, AI Studio is usually a better first stop than Flow, because it exposes the parameters: resolution, the full ratio list including the new 1:4, 4:1, 1:8 and 8:1 at 2K and 4K, up to 14 reference images, and the thinking levels minimal, medium and high.

From a Flow string to general availability, by date

The sequence, because a lot of secondhand coverage compresses it into one event:

• 2026-09-24: a Google Flow web build referenced "Nano Banana 2.5 Flash". A string in the build, not a user-selectable option.
• 2026-09-27: the same string was relabeled "Nano Banana 2.1".
• 2026-10-05: "Nano Banana 2.1" appeared as a selectable option in Flow's image model picker, next to Nano Banana Pro and Nano Banana 2 Lite.
• 2026-10-06: Google released Nano Banana 2.1 and made it generally available, with a model id, a price and a place in several products.

The rename reads differently now than it did at the time. A model going from "2.5 Flash" to "2.1" in three days looked like chaos; in hindsight it was Google settling on describing the model as a point update to Nano Banana 2 rather than as a new half-generation.

For the full picture of what shipped, see what is Nano Banana 2.1, and for how it stacks up against the Pro model, the comparison guide.

If you cannot see Nano Banana 2.1 yet

General availability does not mean everyone sees it at the same moment. Product rollouts are gradual, and which surfaces offer 2.1 depends on your account, your plan and your country. Google has published neither the country list nor the plan limits, so there is no table to check yourself against.

There is no toggle, no beta key and nothing to install that changes it. If Flow is not showing the model, the realistic options are to try another surface, such as the Gemini app, AI Studio or AI Mode, or to use the API, which is available to developers now.

Be careful with anyone selling access. The model is released and priced publicly, so an offer of early access, an unlock or a special key is either a mislabeled different model or a straightforward scam. Do not pay for one and do not paste Google credentials into anything that promises it.

Flow, the API, and what we run here

Flow and the API solve different problems. Flow gives you a project, a canvas and a curated set of controls. The API gives you gemini-nano-banana-2.1, documented parameters, a stated price per image and the ability to build the model into your own tool. Most of the surfaces above are the first kind; only one is the second.

This site is the first kind too: a web generator and editor that reaches the models through a third-party provider's API, independent of Google. Nano Banana 2.1 is the default entry in the prompt box and uses Nano Banana Pro. Nano Banana Pro, Nano Banana 2, Nano Banana 2 Lite and the original Nano Banana are all available in the picker today.

Until then, the workflow that works regardless of surface: draft at 1K on a Flash model to find the framing, then re-render the keeper at 2K or 4K, or on Nano Banana Pro if it needs clean type. Ratios run from 1:1 through 21:9 here, output is 1K, 2K or 4K, credits are pay as you go with credit packs or monthly plans, failed images are refunded, and your library stays private. The prompts guide has prompts written for what 2.1 is good at.

FAQ