Integrate Nano Banana 2.1 through EvoLink's unified image API. Generate or edit at 1K, 2K or 4K, combine up to 14 reference images, and enable search grounding with the same API key.
Playground
Write a prompt, pick a resolution, and see the estimated cost before you spend anything.
Create task
Fixed to gemini-nano-banana-2.1Up to 14 images, 20 MB each. JPEG, PNG or WebP.
Each resolution has its own per-image price.
Search is billed per query the model actually runs. One generation can run multiple queries.
Generated image links stay valid for 24 hours — save anything you want to keep.
Output images are billed by resolution. Reference images, actual search queries, prompt text and thinking tokens are separate usage items.
Google pricing sources · Vertex Standard · Gemini Developer API
One 1K image from a text prompt
About 312 tests with $10 in credits.
This estimate covers one 1K output image with no reference images or search. Prompt and thinking tokens are added from actual usage; the displayed budget counts are estimates for image output only.
Prototyping with the first-test setup above
Batch generation and prompt sweeps
High-volume pipelines and A/B sets
Estimated cost for one generated image. Final cost may vary by output quality and input images.
Image endpoint rates; actual search counts and text/thinking tokens determine the final total.
gemini-nano-banana-2.1Pay for the output tier, each reference image, actual search queries and text/thinking tokens. Estimates assume one query per enabled tool; settlement can be higher. Google comparisons use Vertex Standard pricing; the Developer API lists a different 4K rate.
| Item | Rule | Rate | Billing |
|---|---|---|---|
| Image generation · 1K | The generated image at the 1K tier (about 1 megapixel). One image per request. | $0.031 / image-10% 2.0808 cr / imageofficial Vertex rate $0.034 | Per image |
| Image generation · 2K | The generated image at the 2K tier (about 4 megapixels). One image per request. | $0.046 / image-10% 3.0845 cr / imageofficial Vertex rate $0.051 | Per image |
| Image generation · 4K | The generated image at the 4K tier (about 16 megapixels). One image per request. | $0.103 / image-10% 6.9401 cr / imageofficial Vertex rate $0.114 | Per image |
| Input Image | Applied to each reference image in image_urls, for image-to-image, editing and multi-image composition. | $0.0016 / image-10% 0.1029 cr / imageofficial Vertex rate $0.0017 | Per image |
| Web Search | Billed for each web search query actually executed, not once per generation. Enabled by model_params.web_search. | $0.013 / query-10% 0.8568 cr / queryofficial Vertex rate $0.014 | Per search query |
| Image Search | Billed for each image search query actually executed, not once per generation. Enabled by model_params.image_search. | $0.013 / query-10% 0.8568 cr / queryofficial Vertex rate $0.014 | Per search query |
| Input Text | Produced by the prompt itself. Typically a small share of a request. | $0.0014 / 1K tokens-10% 0.0918 cr / 1K tokensofficial Vertex rate $0.0015 | Per 1K tokens |
| Output Text + Thinking | Text and thinking tokens the model emits alongside the image. Scales with model_params.thinking_level. | $0.0068 / 1K tokens-10% 0.459 cr / 1K tokensofficial Vertex rate $0.0075 | Per 1K tokens |
gemini-nano-banana-2.1Output image price varies by resolution (1K / 2K / 4K); all other enabled items are added to the request
Nano Banana 2.1 is Google's image generation and editing model, released on October 6, 2026, with model ID gemini-nano-banana-2.1. Google reports improvements in visual quality, text rendering and consistency across edits. EvoLink's image task interface combines 1K, 2K and 4K output, reference images and optional search grounding through one API key. Review text, factual content and subject details in your own samples before production use.
Sources checked: October 6, 2026 · Google model documentation · Gemini API release notes
One production route on the image endpoint. Text-to-image and editing share the same model ID — add image_urls to edit.
gemini-nano-banana-2.1Production route. 1K, 2K, or 4K output billed per image by resolution, with optional search grounding and three thinking levels.
Built for everyday image generation and conversational editing at production resolutions.
The fields that shape output and cost on POST /v1/images/generations. Generation is asynchronous — poll the returned task ID or pass a callback, and save results promptly since image links stay valid for 24 hours. Full schemas and code samples live in the API tab.
modelstring · requiredAlways 'gemini-nano-banana-2.1'.
promptstring · requiredThe text instruction describing the image to generate, or how to edit the input images. Up to 4000 tokens. Prompt tokens are metered as text input.
sizestring · default auto'auto' or one of 14 aspect ratios — 1:1, 1:4, 4:1, 1:8, 8:1, 2:3, 3:2, 3:4, 4:3, 4:5, 5:4, 9:16, 16:9, 21:9.
qualitystring · 1K / 2K / 4KSelect 1K, 2K or 4K explicitly; EvoLink defaults to 2K when omitted. The chosen tier sets the output-image rate. This model has no 0.5K output.
image_urlsarray · up to 14Reference image URLs for image-to-image, editing, and multi-image composition. Max 20 MB per image; .jpeg, .jpg, .png or .webp; URLs must be reachable by the server. Each one adds an input-image line item.
model_params.web_searchboolean · default falseNested in model_params. Enables retrieval of current information. Billed by actual web search query count.
model_params.image_searchboolean · default falseNested in model_params. Enables retrieval of visual references. Billed by actual image search query count.
model_params.thinking_levelstring · auto / min / medium / highNested in model_params. Use auto, min, medium or high. Thinking tokens are billed as output text; compare quality and usage on your own samples.
Use the EvoLink API for product integration and batch jobs, or call it from Codex, Claude, or Gemini for fast creative and development workflows. Both paths share the same EvoLink API key, balance, model routes, and task history.
Best for: product backends, batch jobs, automated workflows
Call EvoLink’s unified image API from your server and control the model ID, parameters, task queue, callbacks, and result storage.
Best for: creative and development tasks in Codex, Claude, and Gemini
Give the Agent your output goal, assets, and acceptance criteria. It can choose the route, assemble the request, track the task, and return the result without requiring you to hand-code every step.
Produce 2K and 4K finals for listings, ads, and landing pages from one model.
Retrieve current information for explainers and infographic drafts. Check facts, labels and layout before publishing.
Build sequential edits by passing the previous result as a reference image. Verify that required subject details remain intact.
Create menus, posters and label drafts in the required language and aspect ratio. Review spelling and typography at the final output size.
This example shows the shortest runnable flow: create a task, save the returned task ID, then poll or wait for an HTTPS callback. Open the API tab for the complete parameter and response reference.
curl -X POST https://api.evolink.ai/v1/images/generations \
-H "Authorization: Bearer $EVOLINK_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "gemini-nano-banana-2.1",
"prompt": "A weather infographic for Tokyo this weekend, clean flat design, readable labels",
"size": "16:9",
"quality": "2K",
"model_params": {
"web_search": true
},
"callback_url": "https://your-domain.com/webhooks/image-task-completed"
}'
# Save the returned task ID, then query:
curl https://api.evolink.ai/v1/tasks/{task_id} \
-H "Authorization: Bearer $EVOLINK_API_KEY"Check every value against the parameter cards above: ranges, allowed options, asset counts, and file types must match the selected route.
Check the Authorization bearer token and confirm the available balance and task billing in the console.
Review likeness rights, brand assets, readable text, sensitive content, and input-asset compliance.
A Nano Banana 2.1 task that reaches a final failed state is not charged. Keep the task ID, inspect the final state, and identify whether the cause is validation, review, or execution before retrying.
callback_url must be a public HTTPS URL, not localhost or a private network address. Always keep the task ID as a polling fallback.
On EvoLink, change the model ID, set quality explicitly and retest reference images, search usage and prompts. Google has announced an October 29, 2026 shutdown for gemini-3.1-flash-image on the Gemini Developer API; the retirement date of EvoLink’s existing route must be checked separately.
| Dimension | Nano Banana 2.1 | Nano Banana 2 |
|---|---|---|
| Model ID | gemini-nano-banana-2.1 | gemini-3.1-flash-image-preview |
| Resolution | 1K / 2K / 4K | 0.5K / 1K / 2K / 4K |
| Thinking levels | min / medium / high | min / high |
| Search grounding | Web search and image search | Web search and image search |
| Reference images | Up to 14 | Up to 14 |
| Best for | 1K to 4K workflows with optional search and medium thinking | Existing integrations, or jobs that need 0.5K output |
Send POST /v1/images/generations with model set to gemini-nano-banana-2.1. Text-to-image and editing use the same route — no separate base URL or key.
The create call returns a task ID; poll GET /v1/tasks/{task_id} until it reaches a terminal state, or pass callback_url to be notified on completion, failure, or cancellation. Callbacks to internal IPs are rejected.
Output images are charged by tier, reference images by count, searches by actual query count, and text/thinking by token usage. Failed-task billing follows the task status and settlement policy.
Start at 1K while you iterate, then switch the same request to 2K or 4K for the final — the model ID, prompt, and pipeline stay the same.
Every billable item — output image, reference images, searches, text tokens — is listed separately, and the playground shows an estimate before you generate.
Use one EvoLink key for Nano Banana 2.1 and other image models. Reuse the image task request format, polling and billing interface as you evaluate models for your workflow.
Other Nano Banana image models available on EvoLink.
Nano Banana 2
The previous general-purpose tier, with 0.5K / 1K / 2K / 4K output and search grounding.
View API
Compare resolution, editing features, and per-image cost across the catalogue.
Grok Imagine Image 2.0
xAI image model: text-to-image and image editing on one route, with 1K / 2K output.
View APIModel identity, resolution tiers, search grounding, and billing.
Nano Banana 2.1 is Google’s image generation and editing model, released on October 6, 2026. Its model ID is gemini-nano-banana-2.1. Google reports improved visual quality, typography and consistency across edits. EvoLink’s image task API supports 1K, 2K and 4K output with reference images and optional search grounding.
Google reports improvements in visual quality, text rendering and editing consistency. When switching on EvoLink, use gemini-nano-banana-2.1, select quality explicitly and retest prompts and references. This route adds medium thinking and has no 0.5K output; its default quality is 2K.
Output is available at 1K, 2K, and 4K; 0.5K is not supported. It accepts 14 aspect ratios — 1:1, 1:4, 4:1, 1:8, 8:1, 2:3, 3:2, 3:4, 4:3, 4:5, 5:4, 9:16, 16:9, and 21:9 — or auto.
Output images are billed by resolution tier. Reference images, actual web/image search queries and prompt/thinking tokens are separate usage items. One generation may run multiple search queries. Estimates count one query per enabled tool; the final total uses actual usage. Google price comparisons on this page use Vertex Standard, whose 4K rate differs from the Developer API.
Set model_params.web_search or model_params.image_search to true to retrieve current information or visual references. Both are off by default. Billing follows the number of queries actually executed, not one fixed charge per generation. Review retrieved facts and generated labels before publishing.
Yes. Every generated image carries a SynthID watermark for provenance.
Sign up on EvoLink, create an API key, and send POST /v1/images/generations with model set to gemini-nano-banana-2.1. Generation is asynchronous: poll the returned task ID or pass a callback URL. No Google Cloud account is needed.