Gemini 3.1 Pro Image Preview: API Access & Pricing [2026]

Clearing Up the Names First

Google's image models come with confusing labels, so let's map them before pricing:

• Nano Banana Pro = the high-quality Pro Image model, API ID gemini-3-pro-image-preview. This is what most people mean by "Gemini Pro Image."
• Nano Banana 2 = Gemini 3.1 Flash Image — the cheaper, faster sibling.
• Nano Banana (original) = Gemini 2.5 Flash Image.

If you searched "Gemini 3.1 Pro Image," you're almost certainly after one of two things: the top-quality Pro Image lane (Nano Banana Pro) or the newer 3.1 Flash Image (Nano Banana 2). This guide covers access and pricing for both, with the Pro model as the anchor. All are callable through LinkModel at up to 30% below official rates.

What the Pro Image Model Does

Nano Banana Pro (Gemini 3 Pro Image) is Google's most capable image-generation and editing model. Standout capabilities: ~94% text-rendering accuracy (long passages and multilingual layouts), Search grounding so factual prompts get real geometry and lighting, identity preservation across up to five subjects, natural-language localized edits, 2K/4K output with flexible aspect ratios, and fast 2–12 second generation. Every image ships with a non-optional SynthID watermark — useful for provenance, worth knowing for your compliance story.

Pricing

Google bills per output token, which resolves to clean per-image tiers on the Pro model:

Input is $2.00/1M tokens (text or image) — negligible for simple prompts, meaningful for reference-heavy or high-volume work. Note that 1K and 2K cost the same on the Pro lane because they consume the same output-token amount, so unless you truly need 4K, generate at 2K and save. "Thinking" mode adds a small reasoning-token overhead (5–15% on complex prompts). The 3.1 Flash Image (Nano Banana 2) sits below this on price for higher-volume, cost-sensitive work.

How to Access It

1. Google AI Studio / Gemini API — free tier (around 50–200 images/day depending on surface) to evaluate, then pay-per-use. Use model ID gemini-3-pro-image-preview.
2. Vertex AI — the same family for GCP-native stacks, with Batch/Flex for the async discount.
3. Aggregators — access it through LinkModel alongside GPT Image 2, Seedream, and others behind one API key and one bill.

Calling It Through a Gateway

curl -X POST https://api.linkmodel.ai/api/v1/image-generation \
  -H "Authorization: Bearer $LINKMODEL_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gemini-3-pro-image-preview",
    "prompt": "Infographic titled \"Q3 Growth\" with three labeled bars and a clean legend, corporate blue palette, crisp legible text, 2K"
  }'

The response returns a task_id and the request's price; poll the query endpoint for the finished image. Confirm the exact model string and parameters in the API reference and model catalog, since IDs and availability evolve.

When to Choose It

• Grounded, factual visuals (data viz, real places, product mockups) → Nano Banana Pro's Search grounding is a genuine edge.
• Perfect embedded text at speed → its 94% text accuracy plus 2–12s generation suits high-iteration design.
• Cheaper Google quality at volume → drop to Nano Banana 2 (3.1 Flash Image).
• Reasoning over messy multi-element briefs → compare against GPT Image 2 in Nano Banana Pro vs GPT Image 2.

For the full field and prompting craft, see best AI image generation APIs and the GPT Image 2 prompting guide. For the cost picture across categories, the AI API pricing comparison.

Start free with a $1 credit and test it on your own prompts.