Gemini Pricing [2026 Edition] — Free Tiers, Subscriptions, and API Rates by Model

Gemini pricing falls into three categories. Free tier (Google AI Studio's free API tier, with rate limits and not suited to production), Google subscription (a paid Gemini app plan, with a fixed monthly fee and no API key), and pay-as-you-go API billing (per token, with no monthly fee; $0 if you make no calls). These are separate products: subscribing to one doesn't give you access to another.

Developers are usually looking for the third option. Kunavo offers that API at about 70% below Google's list price. This page covers current rates by model, cost examples you can verify yourself, and the cheapest way to call Gemini in production.

How to distinguish the three pricing options

The most common misunderstanding concerns the middle row. Paying for a Gemini app plan does not increase your API usage allowance. The app's monthly fee and the developer API are billed separately, and to call from code you need a key and pay-as-you-go billing (or the free tier) either way. The free tier's specific rate limits change frequently at Google's discretion, so check the numbers in Google's official documentation. We don't cover them here—if you plan to deploy to production, the free tier isn't an option anyway.

Gemini pricing quick reference

Rates are in USD per 1M tokens, as billed by Kunavo. The “Google list price” column shows Google's published rate for the same model, for comparison.

Flash is the workhorse for high-volume processing; Pro is for more challenging reasoning, image recognition, and long-context jobs. Current rates are always shown on the pricing page and each model page (gemini-2-5-flash, gemini-3-1-pro).

How Gemini token billing works

You're billed for input tokens (everything you send—system prompt, retrieved context, user message) and output tokens (what the model generates). Output costs more, so the biggest lever on a Gemini bill is how much you let the model write. Images and audio are converted to token equivalents and billed on the same meter.

Cost examples

Real figures at Kunavo's Gemini 2.5 Flash rates. The last row alone uses Gemini 3.1 Pro.

At these rates, a batch of 100,000 documents classified with Flash costs about $6, and one million chatbot turns cost about $315. The calculation is executable as written.

# Kunavo の Gemini 2.5 Flash 料率（1M トークンあたり、USD）
IN_RATE, OUT_RATE = 0.09, 0.75

def cost(in_tokens: int, out_tokens: int) -> float:
    return in_tokens / 1_000_000 * IN_RATE + out_tokens / 1_000_000 * OUT_RATE

print(cost(1_000, 300))            # チャットボット 1 ターン -> $0.000315
print(cost(8_000, 500))            # RAG の回答 1 件        -> $0.001095
print(cost(500, 20) * 100_000)     # 10 万件のバッチ処理     -> 約 $6.00

Kunavo pricing and Stripe payments

No subscription and no Google Cloud project required. Top up your balance (Stripe or local payment methods), and each call is deducted at the per-token rates above. Pay as you go, starting with a minimum top-up of $10. The balance never expires, and larger top-ups earn bonus credit. One balance covers Gemini and all other models—Claude, GPT, image, video, audio—so there's no need to reconcile separate invoices for each provider.

Which Gemini model should you choose?

• gemini-2-5-flash — the default for chat, extraction, classification, summarization, and most RAG. It's fast and the cheapest capable option.
• gemini-3-1-pro — use it when Flash isn't accurate enough: multi-step reasoning, code, image recognition, and very long context.

Difficulty-based routing is a good pattern: handle common cases with Flash and escalate to Pro only when a check fails. See the AI cost optimization guide for a routing pattern with code.

How to lower your Gemini bill

1. Use a lower tier. Send the 80% of simple tasks to Flash; save Pro for the 20% of difficult ones.
2. Cap output. Set max_tokens and stop sequences—output is the more expensive side of the meter.
3. Trim input. Retrieve fewer, higher-quality RAG chunks instead of cramming the entire knowledge base into the context.
4. Batch requests. Group independent calls to keep latency low and avoid retry storms.

The same key can also call Google's image and video models. Per-image and per-video pricing is listed in Nano Banana Pro pricing and Veo 3 pricing.

Frequently asked questions