Updated October 8, 2026. Nano Banana 2.1 is officially released. Google has announced its latest image generation and editing model, and the Google Cloud model documentation lists it as generally available with an October 6, 2026 release date and the model ID gemini-nano-banana-2.1.
The earlier Flow menu reports are now release history. The useful questions are what changed, where you can access the model, and how its pricing works.
Nano Banana 2.1 example shared by Google AI Studio in its launch announcement. This is an official example, not a Nano Banana Web test.
Nano Banana 2.1 at a glance
- Release date: October 6, 2026, according to Google's Cloud model page; launch stage: general availability (GA).
- Headline improvements: visual design, mask-based editing, subject consistency, and more natural-looking images, according to Google's announcement.
- Official Cloud model ID:
gemini-nano-banana-2.1. - Documented image output: 1K, 2K, and 4K, with up to 14 input images per prompt.
- Access: Google has announced a rollout across Gemini, AI Mode in Search, Google AI Studio, Flow, Stitch, Google Ads, and Gemini Enterprise Platform.
- Nano Banana Web: Nano Banana 2.1 is available for text-to-image generation and reference-image editing, and is selected by default on its model page.
What is new in Nano Banana 2.1?
Google's launch announcement presents 2.1 as an upgraded image generation and editing model. Google AI Studio's announcement highlights the same improvements and invites users to try it in AI Studio.
Better visual design and more natural-looking images
Google identifies visual design and a more natural appearance as release improvements. For creators, these are useful areas to assess in campaign images, product scenes, portraits, and illustrations: lighting, material detail, composition, and whether the whole image feels coherent.
The launch example above illustrates Google's presentation of the model. It is not a controlled comparison with Nano Banana 2 or Nano Banana Pro, and this guide does not report independent benchmark results.
Mask-based editing
Mask-based editing is one of the announced upgrades. It gives creators a way to direct an edit toward a selected region, such as an object, garment, or background area. That can be useful when a composition is already approved and only one detail needs revision.
The announcement establishes the capability; the exact controls depend on the product or API you use. Do not assume every rollout surface exposes the same mask editor or request format.
Improved subject consistency
Google also highlights subject consistency. This matters when the same person, character, or product appears across several images: the goal is to retain recognizable traits while changing the setting, pose, or lighting.
A practical evaluation is to reuse the same reference images across a short sequence and check faces, clothing, product shape, and other defining details. That is a suggested test workflow, rather than a claim that every sequence will remain identical.
Where can you use Nano Banana 2.1?
In its official rollout follow-up, Google says Nano Banana 2.1 starts rolling out across:
- Gemini
- AI Mode in Google Search
- Google AI Studio
- Flow
- Stitch
- Google Ads
- Gemini Enterprise Platform
A rollout announcement does not mean every account receives the feature immediately. Check the model or image-generation controls in the product you use.
For a direct starting point, visit Google AI Studio. For Cloud access, Google's documentation links to Agent Studio with Nano Banana 2.1 selected and the Model Garden entry. Cloud access requires the relevant project setup and permissions; deploying Google's example app requires billing and the Agent Platform API to be enabled.
Nano Banana 2.1 is available on Nano Banana Web. Open the Nano Banana 2.1 image generator, where it is selected by default. Enter a prompt to create an image, or upload up to 14 reference images to edit an existing visual. Choose an aspect ratio and 1K, 2K, or 4K output; the generator shows the credit cost before you generate.
Official Nano Banana 2.1 specifications
The following are documented on the Google Cloud model page. They describe the Cloud model interface; consumer products may expose different controls or limits.
| Specification | Documented support |
|---|---|
| Model ID | gemini-nano-banana-2.1 |
| Launch stage | General availability (GA) |
| Input | Text, images, and video |
| Output | Text and images; no audio or video output |
| Context window | 131,072 tokens |
| Maximum output | 32,768 tokens |
| Input images per prompt | Up to 14 |
| Output resolutions | 1K, 2K, 4K |
| Aspect ratios | 15 options, including 1:1, 4:5, 9:16, 16:9, and 21:9 |
| Editing | Image editing and multi-turn image editing |
| Grounding | Google Search and image search |
| Additional capabilities | Thinking, system instructions, interleaved text and images, virtual try-on, and C2PA Content Credentials |
| Cloud region | Global |
Video is an input modality: the documentation supports generating images from video input. Nano Banana 2.1 is still an image generation and editing model, not a video-output model.
For developers, Google explicitly marks seed, topK, logprobs, temperature, and topP as unsupported. Sending those parameters returns an API error, so an existing image-model integration needs its request settings checked before switching model IDs.
How much does Nano Banana 2.1 cost?
Google's Cloud pricing page lists these standard global rates, checked for this update:
| Usage | Price per 1 million tokens |
|---|---|
| Input: text, image, and video | $1.50 |
| Cached input | $0.15 |
| Text output, including reasoning | $7.50 |
| Image output | $30.00 |
The model documentation gives the following output-image token counts. At the standard image-output rate, the approximate image-output component is:
| Output resolution | Image output tokens | Approximate output-image cost |
|---|---|---|
| 1K | 1,120 | $0.034 |
| 2K | 1,680 | $0.050 |
| 4K | 3,780 | $0.113 |
These are output-image costs, not the complete request bill. Input images, prompt text, text or reasoning output, and other applicable usage can add charges. Google documents 1,120 tokens per input image. Priority and batch pricing differ; the pricing page says Nano Banana 2.1 does not support Flex.
On Nano Banana Web, Nano Banana 2.1 costs 2 credits per 1K image, 3 credits per 2K image, and 5 credits per 4K image, checked October 8, 2026. The generator displays the cost before generation. These site credit costs are separate from Google Cloud token rates and Gemini subscription limits.
Nano Banana 2.1 vs Nano Banana 2
Nano Banana 2.1 is now a released model with its own documented identity. It should no longer be described only as an unconfirmed Flow label.
Google presents the update as improving visual design, mask-based editing, subject consistency, and natural appearance over its previous models. Those are Google's release claims. We have not yet run matched Nano Banana Web tests to quantify quality or latency differences.
Cloud pricing provides one concrete comparison: the standard global image-output rate is $30 per million tokens for 2.1 versus $60 for Nano Banana 2. Input and text-output rates are higher for 2.1, and 4K output uses more image tokens, so the image-output rate alone does not determine the savings for a complete workflow. Compare the same resolution, reference inputs, and editing sequence.
What happened to the Nano Banana 2.5 Flash name?
Before launch, TestingCatalog reported that a Flow reference changed from Nano Banana 2.5 Flash to Nano Banana 2.1. That explains why older posts and searches mention both names.
The official released name is Nano Banana 2.1, with the Cloud model ID gemini-nano-banana-2.1. The launch sources cited here do not establish whether every earlier 2.5 Flash reference represented the same underlying checkpoint. Use the released name and official documentation when choosing a model or building an integration.
Nano Banana 2.1 FAQ
Has Nano Banana 2.1 officially launched?
Yes. Google and Google AI Studio have announced it, and Google's Cloud documentation lists general availability with an October 6, 2026 release date.
What are the main improvements?
Google highlights visual design, mask-based editing, subject consistency, and more natural-looking images. This guide distinguishes those release claims from independent testing.
Does Nano Banana 2.1 support 4K images?
Yes. The Cloud model documentation lists 1K, 2K, and 4K output resolutions. The controls and limits offered in a particular Google product can differ.
What is the Nano Banana 2.1 API model ID?
Google's Gemini Enterprise Agent Platform documentation lists gemini-nano-banana-2.1. Follow the documentation for the platform you integrate rather than assuming every provider uses the same identifier.
Can I use Nano Banana 2.1 on Nano Banana Web?
Yes. The generator on our Nano Banana 2.1 page selects 2.1 by default and supports text-to-image generation and editing with up to 14 reference images. Choose 1K, 2K, or 4K output and review the displayed credit cost before generating.
Is Nano Banana 2.1 free?
The Cloud API has paid token-based rates. Access, subscriptions, and quotas in Google's consumer products follow each product's own terms. A public launch does not establish unlimited free use.
Update log
- October 8, 2026: Updated Nano Banana Web availability, the default model, supported generation workflows, and site credit costs following the 2.1 integration.
- October 7, 2026: Rewrote the guide following Google's official launch. Added the October 6 GA release date, announced improvements, rollout destinations, Cloud model ID, documented specifications, and standard pricing. Replaced the pre-launch editorial cover with an official launch example.
- October 6, 2026: Added third-party reports of selectable options in Flow and Gemini, with the model-routing uncertainty reported at that time.
- September 28, 2026: Published the original guide about the Flow-build signal and the earlier 2.5 Flash naming trail.