Gemini API Pricing 2026 — Model Costs, Examples, and Cheaper OpenAI-Compatible Access

Gemini is among the best-value options in frontier AI, and Kunavo prices it about 70% below Google’s official list price behind a single OpenAI-compatible API. This guide provides current rates by model, cost calculation examples you can verify yourself, and the cheapest way to call Gemini in production.

Gemini API pricing at a glance

Rates are in USD per 1M tokens, as billed on Kunavo. The “Google list price” column is Google’s published rate for the same model, shown so you can see the difference.

Flash is the workhorse for high-volume workloads; Pro is for harder reasoning, vision, and long-context tasks. Live rates are always shown on the pricing page and each model page (gemini-2-5-flash, gemini-3-1-pro).

How Gemini token pricing works

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

Example cost calculations

Actual figures using Kunavo’s Gemini 2.5 Flash rate, except the last row, which uses Gemini 3.1 Pro:

At these rates, classifying a batch of 100,000 documents on Flash costs about $6, and one million chatbot turns cost about $315. Here’s the runnable calculation:

# Mức giá Kunavo cho Gemini 2.5 Flash (USD trên mỗi 1 triệu token)
IN_RATE, OUT_RATE = 0.09, 0.75

def chi_phi(token_vao: int, token_ra: int) -> float:
    return token_vao / 1_000_000 * IN_RATE + token_ra / 1_000_000 * OUT_RATE

print(chi_phi(1_000, 300))            # một lượt chatbot        -> $0.000315
print(chi_phi(8_000, 500))            # một câu trả lời RAG     -> $0.001095
print(chi_phi(500, 20) * 100_000)     # lô 100k tài liệu        -> ~$6.00

Kunavo pricing and Stripe billing

No subscriptions, no Google Cloud project. You top up a balance (using Stripe or local payment methods), and calls are deducted at the per-token rates above. It’s pay-as-you-go with a $10 minimum top-up; your balance never expires, and larger top-ups earn bonus credits. One balance covers Gemini and every other model — Claude, GPT, image, video, and audio — so you don’t have to reconcile a separate bill for each provider.

Which Gemini model should I choose?

• gemini-2-5-flash — the default for chat, extraction, classification, summarization, and most RAG. Fast and the cheapest capable option.
• gemini-3-1-pro — use it when Flash isn’t accurate enough: multi-step reasoning, coding, vision, and very long context.

A good pattern is to route by difficulty: use Flash for common cases, and upgrade to Pro only when a check fails. See the AI cost optimization guide for a routing pattern in code.

How to cut your Gemini bill

1. Use a lower tier. Send the easy 80% to Flash; save Pro for the hard 20%.
2. Limit output. Set max_tokens and stop sequences — output is the expensive side of the meter.
3. Trim input. Retrieve fewer, better RAG chunks instead of stuffing your entire knowledge base into the context.
4. Batch requests. Group independent calls to keep latency low and avoid retry storms.

Frequently asked questions