Price the image output as image output.
For gemini-nano-banana-2.1, the official pricing table ↗ separates image output from text and thinking. Image output is $30 per million tokens in Standard and $15 in Batch. Using the text output rate would understate this line item.
Token-derived image output rates · square-output reference
| Resolution | Image tokens | Standard / image | Batch / image |
|---|---|---|---|
| 1K | 1,120 | $0.0336 | $0.0168 |
| 2K | 1,680 | $0.0504 | $0.0252 |
| 4K | 3,780 | $0.1134 | $0.0567 |
Derived from the token counts and modality-specific rates in Google’s pricing footnote. Other aspect ratios may use different token counts.
A SEPARATE LINE ITEM
Input: $1.50 / million tokens Standard, $0.75 Batch. Text and thinking output: $7.50 / million Standard, $3.75 Batch. These verified rates are documented here but are not included in the calculator.
Keep the precision until the end.
1,000 × 3,780 × $30 ÷ 1,000,000
$113.40 image output onlyThe provider displays a 4K Standard equivalent of $0.113. Its token formula gives $0.1134. We use the formula, so 1,000 images estimate to $113.40 rather than $113.00. This is an arithmetic estimate, not a claim about invoice rounding.
The calculator uses integer microdollars (one millionth of a dollar), then rounds the combined output estimate once to cents, half-up. The exact amount remains visible and is included in the CSV.
Extra outputs are your assumption.
The optional allowance adds ceil(quantity × percentage / 100) images. For 101 images at 10%, it adds 11 outputs. We assume each additional output is billable; we do not infer whether a failed request, refusal, or retry will be charged.
With a custom rate, the amount you enter applies to the selected mode and resolution as supplied. No Batch discount is applied automatically, and the rate is labeled unverified.
A production budget needs more than output.
Input and conversation history
Prompts, reference images, videos, and repeated context can add input usage. Editing sessions may have a very different input budget from text-only image requests.
Text, thinking, and grounding
Text and thinking output have separate rates. Search can add charges. No search allowance or shared free quota is modeled here; consult the provider’s table for your actual usage.
Output shape and usage
The presets use the square examples in the pricing footnote. Aspect ratio and actual token usage can change the estimate. Count output images rather than API requests; check real usage metadata before committing a large budget.
Your billing arrangement
Estimates exclude tax, currency conversion, credits, negotiated discounts, resellers’ markups, storage, and delivery. API pricing is separate from consumer subscriptions and third-party image tools.
Batch changes the workflow, too.
The Batch API documentation describes asynchronous processing with a 24-hour target turnaround. A target is not a completion guarantee. A lower rate does not make Batch interchangeable with an interactive request.
Traceable rates. Visible updates.
Last verification: 22 Oct 2026 (UTC). Rates are a documented snapshot and are not fetched live. Check the official pages again before a purchasing or migration decision.
- Gemini Developer API pricingRates, units, resolution token counts, and price equivalents ↗
- Gemini API release notesOctober 6, 2026 release and predecessor deprecation ↗
- Nano Banana 2.1 model specificationExact model ID, supported modes, and resolutions ↗
- Batch API guideAsynchronous behavior and target turnaround ↗
Our method, exclusions, and extra output allowance are Framecost’s estimation choices. Model names belong to their respective owners. Framecost is an independent tool.
Build an estimate