Nano Banana Pro pricing & free access: API rates
Quick answer
Nano Banana Pro is Google’s Gemini 3 Pro Image model. In the standard Gemini Developer API, its listed image-output cost is about $0.134 for a 1K or 2K image and $0.24 for a 4K image, plus applicable input and other output charges. Consumer subscriptions and third-party providers use different billing. Checked September 13, 2026 against Google’s API pricing.
The standard Gemini Developer API has no free tier for Nano Banana Pro according to Google’s current pricing table. Google AI Studio usage is free in available regions, while the Gemini app applies separate plan and capacity limits. “Free trial” can therefore mean a browser surface, not free production API calls.
Remove the mug from the right side of the desk. Reconstruct the revealed surface naturally while preserving the laptop, lighting, shadows, camera position, and every other object.
Test an image edit before paying
Upload a representative image, describe one exact change, and inspect the result in Magic Hour.
Try AI Image EditorThis pricing guide uses current Google model and billing documentation and keeps direct API rates separate from hosted-app pricing. It is a price reference rather than a quality ranking; verify the exact model ID, account tier, and final billable route before production use.
Direct API output pricing
Model | Output size | Listed image-output component |
|---|---|---|
1K or 2K | About $0.134/image | |
4K | About $0.24/image |
These are the standard API image-output components, not all-inclusive flat prices. Google lists input at $2 per million tokens for this model and bills applicable text output separately. Batch or Flex modes and other providers can have different prices and behavior.
What would 100 images cost?
At the listed output rate, 100 standard 4K image outputs contribute about $24 in image-output charges, before input and other applicable charges. If you generate two candidates for every accepted image, 100 accepted images require 200 outputs and about $48 in that component.
This is a budget scenario. It does not establish an actual two-attempt success rate or include every item on an invoice.
Pro is not the same as Nano Banana 2
Nano Banana Pro uses Gemini 3 Pro Image. Nano Banana 2 uses Gemini 3.1 Flash Image and has its own rates. The original Nano Banana is another model generation. Record the exact model ID when comparing quality or cost.
Google’s current image-generation guide recommends Nano Banana 2 as the general-purpose choice for balancing performance, price and latency. Nano Banana Pro remains positioned for complex professional assets, advanced localization, brand consistency and precise control. Choose the model for the job rather than treating “Pro” as the automatic default.
Consumer app, API, or third-party access?
Use a consumer app when its interface and allowance suit manual work. Use an API for application integration and usage accounting. A third-party platform may bundle several models with its own credits and limits; its subscription is not Google’s API pricing.
For image creation or editing within a broader media workflow, compare the actual model and quote in Magic Hour’s image tools. Do not infer a universal per-image cost from a vendor’s generic credit balance.
What should you verify before generating?
Confirm the model ID, output size, number of candidates, input assets, billing mode, and provider. Review exact text and product details in every output. Record accepted images separately from successful API jobs, then divide all relevant charges by accepted images.
For historical measured requests through Magic Hour, see our Nano Banana 2 versus Pro benchmark. It reports fifteen paired 1K attempts per model and links the actual output images. Its July 21, 2026 credit observations are not Google's direct-API prices or a current quote for every resolution.
Is Nano Banana Pro free?
For the Gemini Developer API, no: Google lists Nano Banana Pro as unavailable on the free tier. Google AI Studio usage is free in available regions. In the consumer Gemini app, Google’s current access table lists Nano Banana Pro image redo for paid AI plans and warns that limits can change with capacity, testing and availability. Check the specific surface and account before planning a batch.
Frequently asked questions
At the listed output rate, 100 standard 4K image outputs contribute about $24 in image-output charges, before input and other applicable charges. If you generate two candidates for every accepted image, 100 accepted images require 200 outputs and about $48 in that component.
This is a budget scenario. It does not establish an actual two-attempt success rate or include every item on an invoice.
Use a consumer app when its interface and allowance suit manual work. Use an API for application integration and usage accounting. A third-party platform may bundle several models with its own credits and limits; its subscription is not Google’s API pricing.
For image creation or editing within a broader media workflow, compare the actual model and quote in Magic Hour’s image tools. Do not infer a universal per-image cost from a vendor’s generic credit balance.
Confirm the model ID, output size, number of candidates, input assets, billing mode, and provider. Review exact text and product details in every output. Record accepted images separately from successful API jobs, then divide all relevant charges by accepted images.
For historical measured requests through Magic Hour, see our Nano Banana 2 versus Pro benchmark. It reports fifteen paired 1K attempts per model and links the actual output images. Its July 21, 2026 credit observations are not Google's direct-API prices or a current quote for every resolution.
For the Gemini Developer API, no: Google lists Nano Banana Pro as unavailable on the free tier. Google AI Studio usage is free in available regions. In the consumer Gemini app, Google’s current access table lists Nano Banana Pro image redo for paid AI plans and warns that limits can change with capacity, testing and availability. Check the specific surface and account before planning a batch.
