October 6, 2026
Image generation gets cheaper: Nano Banana 2.1 halves the API price per image
Google cut image generation prices but raised input prices: requests with reference images need a separate cost calculation.
🚨 AI News | TestingCatalog
@testingcatalog
GOOGLE 🔥: Nano Banana 2.1 is coming to Google AI Studio, the API and Gemini! > "A model for high-quality image generation and editing: more accurate facts and visual details, better consistency across successive edits, and performance at Flash scale". > Image generation costs $30 per 1,000,000 tokens. > 1K images (1,024 × 1,024 pixels) use 1,120 tokens and cost $0.034 per image. > 2K images (2,048 × 2,048 pixels) use 1,680 tokens and cost $0.050 per image. > 4K images (4,096 × 4,096 pixels) use 2,520 tokens and cost $0.076 per image. Judging by the prices, is this a Pro model!?
Nano Banana 2.1 is live on API >Inputs & Text Generation: Nano Banana 2.1 costs 2.5× to 3× more. > Output(Image): Nano Banana 2.1 is 50% cheaper
· 12.8K views
On Oct 6, Google released Nano Banana 2.1. A 1K image costs $0.0336 through the API, while input token prices have tripled.
For an app that generates images from descriptions, Google has halved the cost of the images themselves compared with Nano Banana 2. When editing with reference images, input is billed separately, so the full request won't necessarily cost half as much.
What got cheaper. Nano Banana 2 previously charged $0.067 for a 1K image. Nano Banana 2.1 now charges $0.0336. Gemini API prices as of Oct 6, 2026:
| What's billed in Standard | Nano Banana 2 | Nano Banana 2.1 | | --- | --- | --- | | 1K image | $0.067 | $0.0336 | | Input, per 1 million tokens | $0.50 | $1.50 |
Google set Nano Banana 2.1 image pricing at $30 per million tokens in Standard and $15 in Batch. Batch costs half as much for batch generation. Text and thinking output in Standard are billed separately at $7.50 per million tokens.
| Image size | Tokens per image | Standard | Batch | | --- | --- | --- | --- | | 1K, 1,024 × 1,024 | 1,120 | $0.0336 | $0.0168 | | 2K, 2,048 × 2,048 | 1,680 | $0.0504 | $0.0252 | | 4K, 4,096 × 4,096 | 2,520 | $0.0756 | $0.0378 |
Google describes the model as an update to Nano Banana 2 with Flash speed and a more economical alternative to Nano Banana Pro. The company claims improved factual accuracy and detail, along with image consistency across successive edits.
The model accepts up to 14 reference images. Google says it can preserve the appearance of up to 4 characters and accurately depict up to 10 objects. In version 2.1, the company also fixed repeating tile artifacts in panoramic formats 1:4, 4:1, 1:8 and 8:1 at 2K and 4K resolutions.
How to get started. You can test the model using the Try in Google AI Studio button on its official model card. To connect it to an app, you need a key from Google AI Studio, the `GEMINI_API_KEY` variable and the library installed with `pip install -U google-genai` for Python or `npm install @google/genai` for JavaScript.
In Python, generation is called with `client.interactions.create(model="gemini-nano-banana-2.1", input="image description")`. The image is returned in `interaction.output_image.data` as base64. Google's documentation includes examples for Python, JavaScript and REST.
There is no free API tier for Nano Banana 2.1. The Paid Tier is activated through Set up billing with a prepayment starting at $5. Google is also rolling out the model in Gemini.
Google will shut down the previous Nano Banana 2, with the ID gemini-3.1-flash-image, on Oct 29, 2026, and instructs users to switch to gemini-nano-banana-2.1.