Nano Banana 2.1 API and Pricing, Explained Through a Real Pipeline

Nano Banana 2.1 is Google's newest image model on the Flash tier, listed on OpenRouter on October 6, 2026, as google/gemini-nano-banana-2.1 and named as the successor to Nano Banana 2. We added it to the Image Generation node and to chat the next day. One model handles text-to-image and image-to-image, it accepts up to 14 reference images in a request, and it renders at a 1K, 2K or 4K setting.

Like the rest of Google's image line, it is not priced per image. It is priced per output image token, and that number only means something once you know how many tokens an image uses. We measured that when we onboarded it, so this post starts there.

The spec sheet

The fourteen ratios run from a 1:8 tall strip to an 8:1 banner, with 9:16 for Stories and Shorts, 4:5 and 5:4 for feeds, and 16:9 and 21:9 for video frames and wide headers in between. The full capability list is on the Nano Banana 2.1 model page.

What Nano Banana 2.1 costs per image

The listed rate as of October 2026 is $0.00003 per output image token, which is $30 per million image tokens. The provider does not publish a token count per resolution, so on October 7, 2026, we rendered the same prompt, a ceramic cup of coffee on a sunlit cafe table, at each setting and read the usage the API returned.

The measured price sits a little above tokens times rate because each response also carried a few hundred text tokens and the prompt itself, both billed at fractions of a cent. Four things stand out.

No setting means 1K. A request with no resolution came back at 1024 pixels with the same token count as an explicit 1K. If you want a larger image, you have to ask for it.

2K is the value step. Doubling each side to 2048 pixels gives four times the pixels for about one and a half times the price. If an image will be cropped, reframed or used as a video first frame, 2K is the setting to start from.

4K is about three times the price of 1K. At $0.116 it is still under a dime and a half for a 4096 pixel square. The render was clean, with fine texture in the glaze and the table grain, and took three times as long.

A reference image costs about a fifth of a cent. A 1K request with one reference image billed $0.037 against $0.036 for a plain prompt. The reference arrives as input tokens, which are cheap. We tested one reference, not fourteen, so treat a heavy multi-reference request as a little more expensive rather than proven.

These counts come from one day of tests on one shape, which is why Treza quotes the measured amounts before a run and then bills what the provider actually charged.

Generate and edit in one model

With a prompt alone, Nano Banana 2.1 generates. With a prompt plus references, it edits or composes. Attach one photo and describe the change, and you get that photo with the change applied. Attach a product, a character and a setting, and describe how they fit together, and it builds one image from all of them, up to 14 inputs.

We ran the single-reference case at onboarding. The reference was a product still of a folded navy linen shirt, and the prompt asked for the same shirt on a white marble counter in a bright studio. It came back as the same shirt, same color, same linen weave, collar and buttons where they were, now resting on veined marble with a window and a plant softly out of focus behind it. That is the edit most product teams want: keep the item, change the set.

It was also fast. The reference edit took about 8 seconds, the same as a plain prompt at 1K. A 2K render took about 13 seconds and a 4K render about 25. That is quick enough to iterate by hand in chat.

How it compares on the same node

Nano Banana 2.1 sits in the Image Generation node's model menu next to the two Google models people most often weigh it against, so the comparison is practical. The prices below are the per-image rates our node holds for each model, as of October 2026: measured at onboarding for 2.1 and measured across production runs for the other two.

The pattern is simple. Nano Banana 2.1 uses the same token counts as Nano Banana 2 at 1K and 2K, at half the rate per token, so it costs about half as much for the same size. At 4K the gap narrows but 2.1 is still the cheaper of the two. Nano Banana Pro is Google's higher tier, priced flat across 1K and 2K and several times the cost of 2.1 at 1K.

What the table cannot tell you is which output you will prefer, because that depends on your prompt. The test is cheap: put the same prompt and references through two models on two nodes and compare. If you are choosing outside Google's line, the Hy Image 3.5 pricing post and the Seedream 5.0 Flash pricing post cover two models that sit in the same menu at different price points.

What you actually pay on Treza

Nano Banana 2.1 through a Treza pipeline is metered from a prepaid credit balance. Treza quotes the measured amounts above, plus its credit markup, before a run. One credit equals $0.01, packs start at $5, credits never expire, and there is no subscription. A failed generation charges nothing, and that includes a prompt the provider refuses.

Calling Nano Banana 2.1 as an API

Direct access means provider credentials, a chat completions request shape that returns images, and your own retry logic before you have built anything.

Through a Treza pipeline, Nano Banana 2.1 is a node on a canvas. Place the Image Generation node, pick the model, set the resolution and aspect ratio, wire references into its reference input, and publish the graph as a single versioned endpoint. Your app calls the typed /invoke route with JSON in and JSON out, or points an existing OpenAI SDK at the OpenAI-compatible route. If you later swap the node to a model from another lab, the endpoint your product calls stays the same.

The same pipeline is reachable from an agent. Treza's hosted MCP server exposes pipelines as tools that Claude and other MCP clients call directly. In chat, asking for Nano Banana 2.1 by name and attaching references does the same thing with no canvas at all.

Putting it in a pipeline instead of a single call

A single image is rarely the deliverable. The more useful pattern is Nano Banana 2.1 as the first node of several.

Compose a key frame from your references at 2K, then pass it to a video node as the first frame. The image to video generator page has that pairing set up, and the post on setting a first-frame image for AI video generation covers the handoff. A fuller AI video pipeline puts a script step ahead of the image and narration, captions, or publishing behind the video, then reruns the whole graph on a schedule or from an API call.

Starting from a blank canvas, the AI image generator page has the image side set up, and the AI video generator page puts a video model behind a single generate action.

Frequently Asked Questions

How much does Nano Banana 2.1 cost per image?

It is listed at $0.00003 per output image token as of October 2026. In our tests on October 7, 2026, a 1K image used 1,120 image tokens and cost $0.036, a 2K image used 1,680 and cost $0.053, and a 4K image used 3,780 and cost $0.116.

Is Nano Banana 2.1 cheaper than Nano Banana 2?

Yes, at every size we compared. It used the same image token counts as Nano Banana 2 at 1K and 2K at half the per-token rate, so it costs about half as much: $0.036 against $0.07 at 1K, and $0.053 against $0.105 at 2K.

Do reference images cost extra with Nano Banana 2.1?

A little. In our test one reference image added about $0.002 to a 1K render, because the image is billed as input tokens. The model accepts up to 14 references per request.

What resolution does Nano Banana 2.1 render?

The node offers 1K, 2K and 4K. A square came back at 1024, 2048 and 4096 pixels for those settings, and at 1024 pixels with no setting at all.

How long does Nano Banana 2.1 take to render?

In our tests, about 8 seconds at 1K, 13 at 2K and 25 at 4K, and about 8 seconds for an edit with one reference image. That was a small sample from one day, so treat it as a rough guide.

Can I use Nano Banana 2.1 in chat?

Yes. Ask for Nano Banana 2.1 by name in chat, attach reference images if you have them, and the image comes back in the conversation.

Can I call Nano Banana 2.1 through an API without integrating Google myself?

Yes. On Treza the model is a node inside a pipeline you publish as a versioned HTTP endpoint, with a typed /invoke route and an OpenAI-compatible route, so your product calls one API whichever image model sits behind the node.