Is Gemini 3.1 Flash Image the same as Nano Banana?

No. Nano Banana refers to Gemini 2.5 Flash Image. Gemini 3.1 Flash Image is a newer preview model that uses the Gemini 3 generation at Flash-tier latency.

Why is it labelled "preview"?

Google ships major image models in preview before GA. Preview models can change behaviour or pricing without the usual deprecation cadence. Pinning a specific model in your PlugNode flow is a one-line config change.

Does Gemini 3.1 Flash Image support image editing?

Yes. Like the rest of the Gemini image family, it accepts a reference image and a text instruction and returns an edited image.

When should I pick this over Nano Banana Pro?

Pick Flash for high-volume routine production. Pick Pro when a brief demands the highest fidelity, fine typography, or precise multi-reference composition.

Can I A/B both models on the same prompt?

Yes. Drop two Image nodes on the canvas, wire the same prompt into both, set each to a different model, and compare the outputs side by side.

Google · Image generation

Gemini 3.1 Flash Image

Google Gemini 3.1 Flash Image on a visual canvas. The balance between the original Nano Banana and the Pro tier, wired into publishable flows.

At a glance

Gemini 3.1 Flash Image is Google's preview-tier image model between Nano Banana and Nano Banana Pro.
Better detail than the original Flash image, faster than the Pro tier.
Selectable on PlugNode's Image node. BYO Gemini API key, no markup.
Good default when you have outgrown Nano Banana but do not need Pro-tier latency.

What it is

Gemini 3.1 Flash Image is Google's preview-tier image model that sits between the original Nano Banana (Gemini 2.5 Flash Image) and Nano Banana Pro (Gemini 3 Pro Image). It uses the new Gemini 3 generation while retaining the Flash performance profile: faster and cheaper than Pro, sharper than the original Flash. On PlugNode it is one of three options on the Image node, selectable from the model dropdown. If you are building a flow that needs better detail than Nano Banana but does not justify Pro-tier costs, this is the default pick.

What you can do with it

Gemini 3 generation at Flash-tier latency
Edit-by-description with better prompt adherence than the 2.5 Flash variant
Multi-reference composition with caption hints
Aspect-ratio control on the node config
Cheaper per-call than Pro for routine production work
Drop-in replacement for Nano Banana via dropdown

Where Gemini 3.1 Flash Image fits

The image-model decision on PlugNode is a three-way pick between the original Nano Banana (speed and cost), Gemini 3.1 Flash Image (balance), and Nano Banana Pro (detail and composition).

Gemini 3.1 Flash Image is the middle option and usually the right default once a flow moves past prototyping. It uses the newer Gemini 3 generation for improved prompt adherence and finer detail, but runs at Flash-tier latency and price. Teams shipping production pipelines tend to start here and only promote to Pro when a specific brief demands it.

What changed between 2.5 Flash and 3.1 Flash

The jump from Gemini 2.5 Flash Image to Gemini 3.1 Flash Image is a generation change, not a revision. Prompt adherence is tighter, small-text rendering is cleaner, and multi-image composition holds up better when the caption hints are specific.

On PlugNode you can run the same prompt through both models inside a multi-model A/B flow: fan out to both Image nodes in parallel and inspect the outputs side by side. That is what the Image node is designed for. The dropdown makes each call cheap to set up. The multi-model A/B testing use case shows the pattern applied to text models; the same shape works for images.

Preview status and what that implies

Gemini 3.1 Flash Image is currently preview in the Gemini API. That means Google can change behaviour or pricing without the deprecation cadence you would expect from a GA model.

In practice, the impact on a shipped PlugNode flow is small. The Image node reads the model ID from config, so a re-pin to a different model version is a config change rather than a code change. The version-history system tracks the change. Roll back if the new version regresses a downstream consumer.

If your flow is revenue-critical and you want the most stable pin, the original Nano Banana (GA) is the safer selection.

A typical pipeline using Gemini 3.1 Flash Image

Image-heavy pipelines tend to use Gemini 3.1 Flash Image as the workhorse and bring in Nano Banana Pro only for final assets.

Example: HTTP Trigger takes a brief. Gemini Text (Gemini 2.5 Flash) drafts a visual direction. Image (Gemini 3.1 Flash Image) generates six variants. Output collects all of them for human review.

Once the winner is picked, a second flow re-runs with Image (Nano Banana Pro) for the final asset. Both flows share the trigger shape and the publish process. Only the model dropdown differs.

See the brand-asset-generator use case for an end-to-end flow that uses this promotion pattern.

Edit-by-description in the middle tier

Gemini 3.1 Flash Image supports the same edit-by-description contract as Nano Banana and Pro: pass a reference image plus an instruction, get back an edit. It sits in the middle on edit quality: better than the 2.5 Flash variant at preserving fine detail during an edit, less exact than Pro.

For high-volume edits where the operator will review outputs and re-run as needed, it is a better tradeoff than calling Pro every time. A flow that publishes an edit endpoint can expose the model as a query parameter so the caller picks the tier at call time.

Run it on PlugNode

Select "Gemini 3.1 Flash Image (Preview)" on the Image node dropdown. Everything else (Gemini API key, aspect ratio, inputs) works the same as the Nano Banana and Pro options. Switching between them is a single config change.

Gemini 3.1 Flash Image is billed by Google at preview-tier Flash rates. PlugNode does not meter generations or add a markup.

Use it inside these workflows

All use cases

Brand designers, marketing teams, startup foundersBrand Asset GeneratorType a brief, generate an on-brand hero image at your target aspect ratio, all in one manual flow.Social media managers, creative ops, agenciesSocial Media Content PipelinePOST a creative brief via webhook, get back a generated hero image for the campaign. Multi-platform sizing is on the roadmap.AI engineers, product teams, prompt designersMulti-Model A/B TestingRun the same prompt through Gemini and OpenAI in parallel. Compare latency, token cost, and output quality on your actual production prompts.

Frequently asked questions

Is Gemini 3.1 Flash Image the same as Nano Banana?: No. Nano Banana refers to Gemini 2.5 Flash Image. Gemini 3.1 Flash Image is a newer preview model that uses the Gemini 3 generation at Flash-tier latency.
Why is it labelled "preview"?: Google ships major image models in preview before GA. Preview models can change behaviour or pricing without the usual deprecation cadence. Pinning a specific model in your PlugNode flow is a one-line config change.
Does Gemini 3.1 Flash Image support image editing?: Yes. Like the rest of the Gemini image family, it accepts a reference image and a text instruction and returns an edited image.
When should I pick this over Nano Banana Pro?: Pick Flash for high-volume routine production. Pick Pro when a brief demands the highest fidelity, fine typography, or precise multi-reference composition.
Can I A/B both models on the same prompt?: Yes. Drop two Image nodes on the canvas, wire the same prompt into both, set each to a different model, and compare the outputs side by side.

Related models

All models

Google · Image generationNano Banana ProGoogle's Gemini 3 Pro Image (the Nano Banana Pro model) on a visual canvas. Generate, edit, and publish image pipelines behind one URL.

Google · Image generationNano BananaGoogle Gemini 2.5 Flash Image (the original "Nano Banana") on a visual canvas. Fast image generation and edits, wired into publishable flows.

Google · Video generationVeo 3.1Google Veo 3.1 on a visual canvas. Generate 4K video with audio, wire it into a pipeline, publish the whole flow as a signed endpoint.

Last updated 2026-04-25

Generate your first video ad in 3 minutes.

Free to start. No credit card. Upload a product photo, connect your AI models, click Run.