Updated

Google

Nano Banana 2.1

Google's Nano Banana 2.1, built on Gemini 3.6 Flash: up to 4K images, clearer text and stronger multi-reference edits.

Create with Nano Banana 2.1View pricing

What is Nano Banana 2.1?

Nano Banana 2.1 is Google DeepMind's image generation and editing model in the Gemini 3 family, built on Gemini 3.6 Flash.

Google published its model card on October 6, 2026. It reads text and images with a context window of up to 1M tokens and returns images up to 4K, with a thinking mode and a faster no-thinking mode.

Google measured it against Nano Banana 2 and Nano Banana Pro in side-by-side human evaluations. In thinking mode it scored an Elo of 1050 on overall text-to-image preference (Nano Banana 2: 990, Nano Banana Pro: 935), 1048 on infographic design (961 and 912) and 1026 on general editing (938 and 939). The largest gap is multi-character consistency, 1106 against 978 and 1011, and on Google's infographic factuality autorater it scored 0.521 against 0.179 and 0.265.

Google lists clear limits too. Small text, long paragraphs and page-length text can blur, especially at 1K. Character consistency is not always perfect, masked and doodle edits can be followed only partly, and left and right can get mixed up. On Flashloop the Nano Banana 2 option now runs Nano Banana 2.1, at 1K, 2K or 4K with up to 10 reference images.

Provider
Google
Resolution
1K, 2K, 4K
Aspect ratios
1:1, 2:3, 3:2, 3:4, 4:3, 4:5, 5:4, 9:16, 16:9, 21:9, 1:4, 4:1, 1:8, 8:1

What you can do with Nano Banana 2.1

  • Built on Gemini 3.6 Flash, with text and image input and a context window of up to 1M tokens

  • Up to 4K output, at 1K, 2K or 4K on Flashloop

  • Clear text for posters, diagrams and infographics, including localized text in several languages

  • Up to 10 reference images on Flashloop for multi-reference edits, characters and products

  • Google Search grounding on Flashloop, so images can draw on real-world facts

Prompts to try with Nano Banana 2.1

“A clean vertical infographic titled "How Coffee Gets to Your Cup" with five numbered steps: Harvest, Process, Roast, Grind, Brew. One flat icon per step, warm brown and cream palette, short label under each icon, 2:3.”

Try this prompt

Where Nano Banana 2.1 fits best

  • Infographics and diagrams

    Google built this version to use Gemini's world knowledge for infographics and diagrams. Give it the facts and the layout and check the small labels before you publish.

  • Posters with clear text

    Headlines and short copy render cleanly. Render at 2K or 4K when the design has small type.

  • Localized text across languages

    Make one design and render versions with the copy in several languages.

  • Multi-reference editing

    Combine up to 10 reference images on Flashloop: a product from one, a person from another, the style from a third.

  • Character and product consistency

    Keep the same character or product across a set of images, the area where Google measured the largest gain over earlier Nano Banana models.

How Nano Banana 2.1 compares

Nano Banana 2.1 compared with other image models on Flashloop
ModelTop resolutionAspect ratiosPlan
Nano Banana 2.14K1:1, 2:3, 3:2 +11All plans
Nano Banana Pro4K1:1, 3:2, 2:3 +2All plans
Flux 24MP (2048x2048)1:1, 3:2, 2:3 +2All plans
Seedream 5.0 Lite4K1:1, 3:2, 2:3 +2All plans
Seedream 5.0 Pro2K (High)1:1, 4:3, 3:4 +5All plans

Good to know before you generate

  • Small text, long paragraphs and page-length text can come out blurry, especially at 1K, per Google's model card
  • Character consistency between input images and the result is not always perfect
  • Masked or doodle-based edits can follow instructions only partly
  • Edits sometimes keep the pose of the input subject when you asked for a change
  • Left and right and other spatial instructions are occasionally confused
  • World knowledge, 3D reasoning and factuality are still limited, and the model can hallucinate details

Start creating with Nano Banana 2.1

Create with Nano Banana 2.1

  1. Select Nano Banana 2.1

    Open Flashloop and choose Nano Banana 2.1 from the model selector in the image creator.

  2. Enter your prompt

    Describe what you want to generate.

  3. Generate and download

    Hit generate and your image will be ready in seconds. Download or share directly.

Questions about Nano Banana 2.1

What is Nano Banana 2.1?

Google DeepMind's image generation and editing model built on Gemini 3.6 Flash. Google published its model card on October 6, 2026. It takes text and images as input and returns images up to 4K.

How does Nano Banana 2.1 compare to Nano Banana 2 and Nano Banana Pro?

In Google's side-by-side human evaluations (Elo scores, thinking mode), Nano Banana 2.1 scored 1050 on overall text-to-image preference, against 990 for Nano Banana 2 and 935 for Nano Banana Pro. On general editing it scored 1026 against 938 and 939, and on multi-character consistency 1106 against 978 and 1011. These are Google's numbers, not Flashloop's.

Is Nano Banana 2.1 better at infographics?

Google reports a large gain: on its infographic factuality autorater, Nano Banana 2.1 in thinking mode scored 0.521, against 0.179 for Nano Banana 2 and 0.265 for Nano Banana Pro. Still check facts and small text before you publish.

How many credits does Nano Banana 2.1 cost on Flashloop?

400 credits at 1K, 600 at 2K and 900 at 4K per image.

What happened to Nano Banana 2 on Flashloop?

The Nano Banana 2 option now runs Nano Banana 2.1, so you get the newer model from the same place in the image creator.

Who makes Nano Banana 2.1?

Google DeepMind. Flashloop offers it alongside other image models; it does not make the model itself.

Does Nano Banana 2.1 have known limitations?

Yes. Google's model card lists blurry small text (especially at 1K), imperfect character consistency, partial instruction following in masked or doodle edits, occasional left and right confusion, and limited 3D reasoning and factuality.

More image models on Flashloop

Browse all models