Back to Models

Gemini 3 Pro Image

googlegoogle/gemini-3-pro-image

Google's GA image generation model built on Gemini 3 Pro. Premium multimodal image-gen model with creative and design capabilities. Replaces gemini-3-pro-image-preview.

Gemini 3 Pro Image, also branded Nano Banana Pro, is Google's GA reasoning-driven image generation and editing model for professional visual work. Google positions it as the strongest Gemini image model for complex and multi-turn generation, image editing, interleaved text-and-image outputs, and prompts that need the model to reason about layout before rendering.

Operationally, it is a text-and-image in/out model with a 65,536-token context window, 32,768 maximum output tokens, up to 14 input images per prompt, 1K/2K output plus 4K preview output, C2PA content credentials, Google Search grounding, batch inference, and Provisioned Throughput support. The model does not support audio, video, Gemini Live API, function calling, structured output, tuning, or OpenAI chat completions on Google's native surface.

Through TheRouter, Gemini 3 Pro Image should be treated as a premium async image model: call the OpenAI-compatible Images API path with model: "google/gemini-3-pro-image", submit with ?async=true, then poll the returned job. Google prices the standard endpoint at $2/M input tokens, $12/M text output tokens, and $120/M image output tokens; 1K and 2K images consume 1,120 image output tokens, while 4K consumes 2,000.

Best for
  • β€’ Professional image generation where prompt reasoning matters β€” posters, campaign hero art, product visuals, technical diagrams, and visual concepts with many constraints
  • β€’ Multi-turn image editing β€” iterative design reviews where a source image is refined over several instructions rather than regenerated from scratch
  • β€’ Reference-heavy creative workflows β€” up to 14 input images per prompt for style, product, character, or layout references
  • β€’ Enterprise image pipelines needing C2PA credentials, global region access, batch inference, and Provisioned Throughput rather than hobby-grade interactive tools
Reach for something else if
  • β€’ Cheap draft generation β€” image output is $120/M tokens ($0.134 for 1K/2K, $0.24 for 4K at Google list price); use Gemini 3.1 Flash Image or a lower-cost image model for bulk ideation
  • β€’ Realtime or synchronous user flows β€” image jobs can outlive edge request limits, so production integrations should use TheRouter's async submit-and-poll pattern
  • β€’ Tool-calling agents, JSON extraction, embeddings, audio, or video β€” the model is specialized for text/image generation and editing, not general chat-agent APIs
Context Length
66K
Max Output
33K
Input Priceper 1M tokens
$2.19/ 1M tokens
Output Priceper 1M tokens
$131.65/ 1M tokens

Modalities

textimage→textimage

Capabilities

VisionImage Generation

Pricing Breakdown

TypeRate
Input$2.19 / 1M tokens
Output$131.65 / 1M tokens

Single blended output rate = the image-output rate ($120/MTok). Google also publishes a separate text/thinking output rate of $12/MTok that this schema cannot express, so non-image output tokens are over-billed. Source: ai.google.dev/gemini-api/docs/pricing, read 2026-07-29.

Supported Parameters

temperaturemax_tokenstop_presponse_formatstop

Specifications

Launch stageGAdocs.cloud.google.com β†—verified
Release date2026-05-28docs.cloud.google.com β†—verified
Retirement date2027-05-28 or laterdocs.cloud.google.com β†—verified
Context window65,536 tokensdocs.cloud.google.com β†—verified
Maximum output tokens32,768 tokensdocs.cloud.google.com β†—verified
ModalitiesText and image input/output; audio and video not supporteddocs.cloud.google.com β†—verified
Maximum input images14 images per promptdocs.cloud.google.com β†—verified
Output resolutions1K, 2K, 4K (4K is Preview)docs.cloud.google.com β†—verified
Supported aspect ratios1:1, 3:2, 2:3, 3:4, 4:3, 4:5, 5:4, 9:16, 16:9, 21:9docs.cloud.google.com β†—verified
Supported image MIME typesimage/png, image/jpeg, image/webp, image/heic, image/heifdocs.cloud.google.com β†—verified
Input size limit500 MBdocs.cloud.google.com β†—verified
Input image accounting560 input image tokens per input imagedocs.cloud.google.com β†—verified
Image output accounting1,120 output tokens for 1K/2K images; 2,000 output tokens for 4K imagescloud.google.com β†—verified
Standard Google list price$2/M input tokens; $12/M text output tokens; $120/M image output tokenscloud.google.com β†—verified
CapabilitiesThinking, system instructions, image generation/editing, interleaved images and text, C2PA, Count Tokens, Google Search groundingdocs.cloud.google.com β†—verified
Unsupported capabilitiesGemini Live API, structured output, context caching, RAG Engine, native chat completions, tuning, URL context, code execution, function callingdocs.cloud.google.com β†—verified
LicenseProprietary (Google)verified

Benchmarks

BenchmarkDistributionScoreSource
Google model card β€” complex image generation and editing
Google's public spec describes Gemini 3 Pro Image as the best model for complex and multi-turn image generation and editing, with improved accuracy and enhanced image quality, but does not publish a numeric benchmark table on the model page.
Best Gemini image model for complex and multi-turn image generation/editing; numeric benchmark not publicly discloseddocs.cloud.google.com β†—
Output-resolution support
Resolution support is a production-relevant capability benchmark for routing: 1K and 2K are supported on the GA endpoint, while 4K remains preview and should be guarded behind retry/fallback logic.
1K, 2K, 4K previewdocs.cloud.google.com β†—

API Usage Examples

Use the global api.therouter.ai endpoint shown below for new integrations; the legacy China accelerated endpoint is retired.

Recommended: use the async API

Image generation typically takes 30–180s, beyond the edge sync timeout. The examples below use the ?async=true submit + poll pattern. Read the full async image generation & edit guide β†’

cURL
# 1) Submit job (returns 202 immediately with a polling URL).
# Image generation takes 30-180s β€” always use the async path in production.
JOB=$(curl -s -X POST "https://api.therouter.ai/v1/images/generations?async=true"   -H "Content-Type: application/json"   -H "Authorization: Bearer $THE_ROUTER_API_KEY"   -d '{
    "model": "google/gemini-3-pro-image",
    "prompt": "A cinematic product render with soft studio lighting"
  }' | python3 -c "import sys,json;print(json.load(sys.stdin)['id'])")
echo "submitted: $JOB"

# 2) Poll until terminal (succeeded / failed / cancelled / expired).
while :; do
  R=$(curl -s "https://api.therouter.ai/v1/jobs/$JOB"     -H "Authorization: Bearer $THE_ROUTER_API_KEY")
  S=$(echo "$R" | python3 -c "import sys,json;print(json.load(sys.stdin)['status'])")
  echo "status: $S"
  case "$S" in
    succeeded) echo "$R" | python3 -c "import sys,json;print(json.load(sys.stdin)['unsigned_urls'][0])"; break ;;
    failed|cancelled|expired) echo "$R"; exit 1 ;;
  esac
  sleep 5
done

Async image generation

Submit image-generation jobs through TheRouter's OpenAI-compatible Images API. Use the async path for production so a long-running render does not hit request timeouts.

cURL
# 1) Submit an async generation job.
JOB=$(curl -s -X POST "https://api.therouter.ai/v1/images/generations?async=true"   -H "Authorization: Bearer $THEROUTER_API_KEY"   -H "Content-Type: application/json"   -d '{
    "model": "google/gemini-3-pro-image",
    "prompt": "Create a clean 16:9 product hero image for an AI routing dashboard, dark background, teal highlights, crisp UI typography.",
    "size": "1792x1024",
    "n": 1
  }' | python3 -c "import sys,json;print(json.load(sys.stdin)["id"])")

# 2) Poll the job until it succeeds or fails.
while :; do
  R=$(curl -s "https://api.therouter.ai/v1/jobs/$JOB"     -H "Authorization: Bearer $THEROUTER_API_KEY")
  S=$(echo "$R" | python3 -c "import sys,json;print(json.load(sys.stdin)["status"])")
  echo "status: $S"
  case "$S" in
    succeeded) echo "$R" | python3 -c "import sys,json;print(json.load(sys.stdin)["unsigned_urls"][0])"; break ;;
    failed|cancelled|expired) echo "$R"; exit 1 ;;
  esac
  sleep 5
done

Image Editing Examples

Upload an image and describe the edit you want with a text prompt; the model returns the edited image as base64.

cURL
# Same async submit + poll pattern as /v1/images/generations.
JOB=$(curl -s -X POST "https://api.therouter.ai/v1/images/edits?async=true"   -H "Authorization: Bearer $THE_ROUTER_API_KEY"   -F "model=google/gemini-3-pro-image"   -F "prompt=Turn this scene into a watercolor painting"   -F "size=1024x1024"   -F "image=@input.png" | python3 -c "import sys,json;print(json.load(sys.stdin)['id'])")
echo "submitted: $JOB"

while :; do
  R=$(curl -s "https://api.therouter.ai/v1/jobs/$JOB"     -H "Authorization: Bearer $THE_ROUTER_API_KEY")
  S=$(echo "$R" | python3 -c "import sys,json;print(json.load(sys.stdin)['status'])")
  echo "status: $S"
  case "$S" in
    succeeded) echo "$R" | python3 -c "import sys,json;print(json.load(sys.stdin)['unsigned_urls'][0])"; break ;;
    failed|cancelled|expired) echo "$R"; exit 1 ;;
  esac
  sleep 5
done

More from google

Similar models

Cross-provider sibling models

News & changes

2026-05-28

Gemini 3 Pro Image reaches GA as Nano Banana Pro

Google's Agent Platform model page lists Gemini 3 Pro Image as a GA endpoint with a May 28, 2026 release date, global availability, reasoning support, image generation/editing, C2PA content credentials, batch inference, and a support floor through at least May 28, 2027.

re-authored by TheRouterdocs.cloud.google.com β†—
2026-05-28

Google publishes Gemini 3 Pro Image token pricing

Google Cloud pricing lists the standard Gemini 3 Pro Image endpoint at $2/M input tokens, $12/M text output tokens, and $120/M image output tokens, with 1K/2K images billed as 1,120 output tokens and 4K as 2,000 output tokens.

re-authored by TheRoutercloud.google.com β†—

Frequently asked

Is Gemini 3 Pro Image available through an OpenAI-compatible API on TheRouter?

Yes. Use TheRouter's OpenAI-compatible image endpoints with model: "google/gemini-3-pro-image". For production, prefer POST /v1/images/generations?async=true or /v1/images/edits?async=true, then poll the returned job URL.

re-authored by TheRoutertherouter.ai β†—
How much does Gemini 3 Pro Image cost per image?

At Google list price, image output is $120 per 1M image output tokens. Google states 1K and 2K outputs consume 1,120 tokens, or about $0.134 per image, while 4K outputs consume 2,000 tokens, or about $0.24 per image. Text and input-image tokens are billed separately.

re-authored by TheRoutercloud.google.com β†—
What image inputs and output sizes does Gemini 3 Pro Image support?

Google lists up to 14 input images per prompt, PNG/JPEG/WebP/HEIC/HEIF inputs, supported aspect ratios including 1:1, 16:9, 9:16, and 21:9, and output resolutions of 1K, 2K, and 4K. 4K output is still marked Preview.

re-authored by TheRouterdocs.cloud.google.com β†—
Does Gemini 3 Pro Image support function calling or structured JSON output?

No. Google's public model spec marks function calling and structured output as not supported. Route tool-calling agents, JSON extraction, and schema-constrained generation to Gemini 3.5 Flash, Gemini 2.5 Pro, or another chat model instead.

re-authored by TheRouterdocs.cloud.google.com β†—
Should I use Gemini 3 Pro Image or GPT Image 2?

Use Gemini 3 Pro Image when Google Search grounding, C2PA credentials, Google Cloud governance, many reference images, or Gemini-family routing consistency matter. Use GPT Image 2 when your team wants the most OpenAI-native image API behavior or already standardized prompts around OpenAI image models. In either case, run a small brand-specific eval set before moving bulk generation.

re-authored by TheRouterdocs.cloud.google.com β†—
Fact ledger β€” every claim on this page traces here
sourceURLretrieved
Launch stagedocs.cloud.google.com β†—2026-07-25verified
Release datedocs.cloud.google.com β†—2026-07-25verified
Retirement datedocs.cloud.google.com β†—2026-07-25verified
Context windowdocs.cloud.google.com β†—2026-07-25verified
Maximum output tokensdocs.cloud.google.com β†—2026-07-25verified
Modalitiesdocs.cloud.google.com β†—2026-07-25verified
Maximum input imagesdocs.cloud.google.com β†—2026-07-25verified
Output resolutionsdocs.cloud.google.com β†—2026-07-25verified
Supported aspect ratiosdocs.cloud.google.com β†—2026-07-25verified
Supported image MIME typesdocs.cloud.google.com β†—2026-07-25verified
Input size limitdocs.cloud.google.com β†—2026-07-25verified
Input image accountingdocs.cloud.google.com β†—2026-07-25verified
Image output accountingcloud.google.com β†—2026-07-25verified
Standard Google list pricecloud.google.com β†—2026-07-25verified
Capabilitiesdocs.cloud.google.com β†—2026-07-25verified
Unsupported capabilitiesdocs.cloud.google.com β†—2026-07-25verified
Licenseβ€”β€”verified
Google model card β€” complex image generation and editingdocs.cloud.google.com β†—2026-07-25verified
Output-resolution supportdocs.cloud.google.com β†—2026-07-25verified
Gemini 3 Pro Image reaches GA as Nano Banana Prodocs.cloud.google.com β†—2026-07-25verified
Google publishes Gemini 3 Pro Image token pricingcloud.google.com β†—2026-07-25verified
Is Gemini 3 Pro Image available through an OpenAI-compatible API on TheRouter?therouter.ai β†—2026-07-25to verify
How much does Gemini 3 Pro Image cost per image?cloud.google.com β†—2026-07-25to verify
What image inputs and output sizes does Gemini 3 Pro Image support?docs.cloud.google.com β†—2026-07-25to verify
Does Gemini 3 Pro Image support function calling or structured JSON output?docs.cloud.google.com β†—2026-07-25to verify
Should I use Gemini 3 Pro Image or GPT Image 2?docs.cloud.google.com β†—2026-07-25to verify
Help & contact