Nano Banana API image output costs $0.039 per image up to 1024 × 1024 pixels at Google's standard Gemini Developer API rate, with input and applicable text-output charges billed separately. The original model is Google's hosted Gemini 2.5 Flash Image; use gemini-2.5-flash-image for the request below. Pricing, model documentation
What are the key facts about the Nano Banana API?
| Fact | Verified detail |
|---|---|
| Developer | Google DeepMind. Launch announcement |
| Released | August 26, 2025, initial public release. Launch announcement |
| Type | Image and text inputs; image and text outputs. Model documentation |
| Size or parameters | Not published in the model documentation. Model documentation |
| License and access | Hosted service under Google's API terms, with no open weights. Launch announcement, terms |
| Where it runs | Google-hosted inference accessed using an API client or Google's interfaces. Launch announcement |
The identifier for the original model is gemini-2.5-flash-image. Its model page lists the older preview identifier as deprecated. Use the documented stable name when starting a new integration. Model documentation
What can you build with the Nano Banana API?
The original Nano Banana accepts reference images alongside written instructions. Google documents editing, combining images, and preserving a subject across different scenes as intended capabilities. Launch announcement
That interface supports a useful application pattern: let a user supply a reference, describe the requested transformation, then inspect a returned image.
You can prototype the task in AI Studio before deciding how your own interface should collect those inputs.
Keep the first integration narrow. For example, build a test that creates a single still life from a text description. Once that succeeds, add image input and evaluate whether the result respects a simple edit instruction.
Record success in terms of the image you actually need. An HTTP response alone does not establish that the image has the correct subject, composition, or lettering. Keep those visual acceptance checks separate from request handling.
What should you check before integrating the Nano Banana API?
Nano Banana access does not include downloadable weights or a local GPU installation.
Your application calls Google's hosted model; local hardware specifications do not establish its service response time. Launch announcement, model documentation
Google's image guide notes that output counts may differ from the number requested.
Build your result handling around the parts actually returned instead of assuming that every successful response contains exactly the output your interface expected. Image guide
The original model also lacks search grounding according to its capability table. If the task requires information from a current web search, choosing this model does not add that feature to the request. Model documentation
Treat API billing as a separate planning task.
A consumer image-generation experience and a developer API price are different access contexts; this article's prices refer specifically to the Gemini Developer API pricing table. Pricing
How much does the Nano Banana API cost and how do you call it?
How do you estimate Nano Banana API costs?
Create a Gemini API key through AI Studio and associate it with the appropriate Google Cloud project. Google's key documentation explains project and key management.
Keep the key in the server environment when building an application. API key guide
As checked on September 5, 2026, Google's standard pricing lists $0.039 per output image up to 1024 × 1024 pixels for Gemini 2.5 Flash Image. That price corresponds to 1,290 output tokens at $30 per million image-output tokens; text and image input costs $0.30 per million tokens.
The table lists no free API tier for this model. Pricing
The output figure is not an all-inclusive project budget. Include input charges, any applicable text-output charges, and additional generations in your estimate.
Follow the pricing page's modality-specific details rather than multiplying an advertised image price and stopping there. Pricing
For a pilot, choose a spending ceiling and a fixed set of representative prompts. Track how many outputs you accept and how many need another attempt. This produces a useful cost-per-accepted-image estimate for your own workload.
How do you send a Nano Banana API request?
Set GEMINI_API_KEY in your environment, then use the following generateContent request. The model identifier comes from the model page; the payload follows Google's API reference. Model documentation, API reference
curl --fail-with-body \
"https://generativelanguage.googleapis.com/v1beta/models/gemini-2.5-flash-image:generateContent" \
-H "x-goog-api-key: $GEMINI_API_KEY" \
-H "Content-Type: application/json" \
-d '{"contents":[{"parts":[{"text":"Create a product photograph of a plain green notebook on a wooden desk."}]}],"generationConfig":{"responseModalities":["TEXT","IMAGE"]}}' \
-o response.json
Read response.json as structured data. Inspect candidate content parts, decode image inlineData.data from base64, and use its mimeType to choose the output format.
Preserve any explanatory text as a separate field. API reference
Handle a response with no image before displaying a success state. Inspect the error or completion information and let the caller know what happened. Do not silently substitute an older image from an earlier request.
How do you add a reference image to the request?
The documented content format accepts image data together with text. Add a reference image only after the basic request works, and make the first editing instruction simple enough to judge visually. API reference
A useful test instruction is to change the notebook cover to blue while preserving its position and the desk. This is an original example, not a tested performance claim.
Compare the changed cover and unchanged areas before accepting the result.
For more detailed editing prompts and a reference-image example, continue with the sibling Nano Banana photo transformation guide.
How does the Nano Banana API differ from Pro?
| Choice | What changes for the integration |
|---|---|
| Nano Banana | Uses gemini-2.5-flash-image; supports image editing without search grounding. Model documentation
|
| Nano Banana Pro | Uses gemini-3-pro-image; the model page documents thinking and search grounding. Pro documentation
|
For interactive access and the distinction between app and API billing, see Nano Banana online access.
What are common Nano Banana API pricing and setup questions?
What is the Nano Banana API model name?
For the original Nano Banana API, use gemini-2.5-flash-image. Google's model page distinguishes that stable name from the deprecated preview identifier. Model documentation
Is the Nano Banana API free?
Google's current pricing table lists no free API tier for Gemini 2.5 Flash Image. The standard output-image price is only one component of the bill; consult input and other applicable charges too. Pricing
Do I need a GPU to call Nano Banana?
The Nano Banana API sends requests to Google's hosted infrastructure. It does not require loading Nano Banana weights on your own GPU. Launch announcement
Why did my request return JSON instead of an image file?
generateContent returns a structured response with candidate content parts. Extract and decode the image data instead of renaming the JSON file to an image extension. API reference
Sources
- Google: Introducing Gemini 2.5 Flash Image
- Google: Gemini 2.5 Flash Image model documentation
- Google: Gemini Developer API pricing
- Google: Using Gemini API keys
- Google: GenerateContent API reference
- Google: Image generation guide
- Google: Gemini API service terms
- Google: Gemini 3 Pro Image model documentation
Top comments (0)