Gemini Api Pricing: Complete 2026 Cost Guide

The Gemini API is billed per token. Prices range from $0.10 per 1 million input tokens on Gemini 2.5 Flash-Lite to $18 per 1 million output tokens on Gemini 3.1 Pro Preview past a 200,000-token prompt. The flagship Gemini 3.1 Pro Preview costs $2 per 1 million input tokens and $12 per 1 million output tokens under that threshold.

Our team at Cabina.AI tracks pricing for Google directly, since we give developers access to Gemini 3.1 Pro Preview alongside 55+ other models through one subscription – when a rate changes, we see it firsthand. By our observation, Google spreads its current lineup across more active model families than either OpenAI or Anthropic, so keeping this page current takes more tracking, not less.

Gemini API Pricing Table – All Models

Google bills the Gemini API per token, not per request or per conversation. The table below breaks down current gemini api pricing for every generally billable model, in US dollars per 1 million tokens, including the previous generation still running in production.

Gemini 3.6 Flash and 3.7 Flash share one rate card with 3.8 Flash, the newest of the group. Each is a separate model ID shipped at a different point in 2026, not a repeated name for the same checkpoint, and Google has not said how long the shared pricing will hold once a newer Flash version ships alone.

Gemini 2.5 Flash-Lite is the cheapest model on the entire price list, current or legacy, at $0.10 input and $0.40 output. It is easy to miss this if you only check the 3.1 Flash-Lite row: the newer “Lite” model actually costs more per token than the one it is meant to replace, at $0.25 input against $0.10.

Vertex AI and the Gemini Enterprise Agent Platform, Google own enterprise channels, price the same model names under separate rate cards. The numbers above cover the first-party Gemini API, billed through Google AI Studio and a standard API key.

A Pricing Cliff Coming on January 1, 2027

Gemini 3.8 Flash is billed at a promotional rate through December 31, 2026: $0.75 input and $3.75 output. The 3.6 Flash and 3.7 Flash releases carry that same promotional rate. Standard pricing doubles to $1.50 input and $7.50 output on all three starting January 1, 2027. Anyone budgeting Flash usage for next year should plan around the higher number, not the current one – the discount is temporary and Google has published the exact date it ends.

Gemini API Cost Calculator

Pick a model and enter your expected monthly token volume below to estimate a monthly bill, with or without the Batch API discount.

Gemini API Request Example

The example below sends one authenticated request to Gemini 3.8 Flash through the current REST endpoint. Google own quickstart uses a simplified input field rather than the older contents array, so swap the model field for any model ID from the table above to price a different model.

curl -X POST "https://generativelanguage.googleapis.com/v1beta/interactions" \
  -H "x-goog-api-key: $GEMINI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gemini-3.8-flash",
    "input": "What should I search for to find the latest developments in renewable energy?"
  }'

The same request in Python, using the official google-genai SDK:

from google import genai

client = genai.Client(api_key="your-api-key-here")

response = client.models.generate_content(
    model="gemini-3.8-flash",
    contents="What should I search for to find the latest developments in renewable energy?",
)

print(response.text)

Model IDs to use in place of gemini-3.8-flash: gemini-3.1-pro-preview, gemini-3.5-flash-lite, gemini-3.1-flash-lite, and gemini-2.5-flash-lite.

Pricing Tiers for Google, Explained

Batch API (-50%)

Batch requests process asynchronously, usually within 24 hours, and both input and output tokens cost 50 percent less than standard pricing. Batch pricing is the simplest way to cut gemini api cost for anything that is not time-sensitive, like bulk classification or offline summarization.

Long-context pricing (over 200,000 tokens)

Gemini 3.1 Pro Preview doubles its rate once a request crosses 200,000 tokens of combined prompt and output: input rises from $2.00 to $4.00, and output from $12.00 to $18.00, per 1 million tokens. Google older 2.5 Pro model uses the same doubling pattern at the same 200,000-token threshold.

Google Search grounding

Connecting a Gemini 3.x model to live Google Search results is free for the first 5,000 requests each month, shared across every 3.x model on the account. Past that, Google charges $14 per 1,000 grounded requests, on top of standard token pricing for whatever the model reads and writes. Gemini 2.5 models use a separate, older grounding price of $35 per 1,000 grounded prompts after 1,500 free requests per day.

Fine-tuned model inference (+50% on Gemini 3 and later)

Supervised fine-tuning itself bills by training token, not by a flat fee. The detail that catches people off guard sits on the inference side: for Gemini 3 and later, a tuned model’s endpoint predicts at 1.5 times the base model’s per-token price. Tuned versions of Gemini 2.5 and earlier models still predict at the ordinary base rate, so the cost of fine-tuning changed materially between generations.

Usage tiers (Free, Tier 1, Tier 2, Tier 3)

Every project starts on the Free tier, with no billing enabled and the lowest per-minute request caps. Enabling billing unlocks Tier 1. Spending $100 over at least three days unlocks Tier 2, and spending $1,000 over at least thirty days unlocks Tier 3, raising the per-minute request and token caps for every model, along with a separate daily request cap. The published per-token price does not change between tiers – what changes is how much of that price a project is actually allowed to spend per minute before Google returns a rate-limit error.

Gemini API Pricing Beyond the Chat Models

The pricing table above covers chat and reasoning models only. Gemini Embedding 2, the current embeddings model, bills separately and far below chat-model rates: $0.20 per 1 million tokens for text, $0.45 per 1 million for image input, $6.50 per 1 million for audio, and $12.00 per 1 million for video. A team that only checks the chat-model table can badly underestimate a bill that also runs embeddings or a retrieval pipeline alongside it.

Function calling and other tool definitions add to the input token count on a request rather than showing up as a separate line item – a tool schema and its arguments count as ordinary input tokens, billed at the model’s standard rate. The Gemini API also supports connecting external tools through the Model Context Protocol, billed the same way: as input and output tokens on whichever model handles the call.

Google own coding-agent tooling, including Gemini CLI and Antigravity, runs on its own plans and rate cards, separate from the base Gemini API covered here. Anyone pricing out a coding agent specifically should check that product’s own pricing page rather than the model rates above.

Gemini Context Caching and Volume Discounts

Context caching (storing a long prompt so the model does not reprocess it on every call) cuts a cached input token to roughly 10 percent of the standard input price on every current Flash and Pro model, matching the figures in the pricing table above. Google adds a separate storage fee on top, priced per 1 million cached tokens per hour the cache stays alive, typically around $1.00 depending on the model.

The short answer to how much do gemini tokens cost once caching enters the picture is: much less, provided the same long context – a system prompt or a large reference document – gets reused across enough calls to cover the storage fee.

Google does not publish a self-serve volume discount tier for the direct Gemini API. Large-scale and enterprise customers negotiate custom rates through Google Cloud sales, typically by moving usage onto Vertex AI or the Gemini Enterprise Agent Platform.

Which Gemini Model Should You Use?

Gemini 2.5 Flash-Lite or Gemini 3.1 Flash-Lite is the right starting point for high-volume, low-complexity work: classification, extraction, short chat turns, anything where latency and per-call cost matter more than depth of reasoning. Comparing gemini models api pricing side by side, the Lite tier is where most of the token volume in a production app should land.

The 3.x Flash lineup covers the middle of the pricing table, built for general-purpose agents and everyday coding help at a fraction of the Pro rate. In our view, the shared rate card across those Flash releases makes this an easy default: there is no pricing penalty for picking whichever one ships the feature you need.

Gemini 3.1 Pro Preview costs six times as much as Flash on input and stays the model to reach for on long-horizon agentic tasks or genuinely hard reasoning problems. From our experience running Gemini at Cabina.AI, teams tend to reach for Pro Preview by default and only discover the Flash lineup handles their workload just as well after the first invoice arrives.

Is the Gemini API Free?

Partly. Google AI Studio access is free of charge in every supported region, and the free tier includes limited access to select models with no token charge – in exchange, that traffic can be used to improve Google products, unlike paid-tier traffic. Once an account moves to the paid tier and calls a model through a standard API key, every token is billed at the rates in the table above; Google does not publish a separate signup credit the way some competitors do.

Google AI Pro/Ultra vs Gemini API: Which Is Cheaper?

Google AI Pro is the mainstream consumer subscription for using Gemini, priced at $19.99 a month. It covers the Gemini app and Google Search directly, plus Gemini features inside Workspace. Google AI Ultra sits above it as a higher-usage consumer tier for people who need the largest Gemini quotas and extra storage, priced well above Pro.

Both consumer plans meter usage in app-level quotas, not tokens, and neither one includes programmatic API access. For a person chatting with Gemini directly, a subscription is usually the simpler and cheaper option. For anything built to call a model from code, at variable volume, the Gemini API is the only route in, and its bill scales with actual token usage instead of a flat monthly fee.

If you would rather not commit to a single provider before you decide, Gemini is also available on Cabina.AI alongside ChatGPT and Claude. The subscription covers 55+ models in total from one account, which is a lower-commitment way to compare outputs before committing to an API integration. See how Gemini pricing compares to ChatGPT and Claude pricing for a fuller breakdown of the other two.

Gemini API vs ChatGPT API Pricing

Gemini api pricing per million tokens undercuts the ChatGPT API at the cheap end of each lineup, and lands close to it at the flagship tier.

Use the calculator above to price a specific Gemini model against your own token volume. We cover the full breakdown of every OpenAI model in a separate ChatGPT API Pricing guide on this blog. From running both APIs side by side at Cabina.AI, the gap that actually shows up on an invoice is the Pro Preview versus GPT-6 Astra flagship rate – Gemini prices its top model at less than half the ChatGPT flagship rate, before either provider’s long-context or reasoning-effort premiums apply.

Gemini API Price and Model History

Gemini 3 Pro launched in November 2025 as a preview model, then was retired once Gemini 3.1 Pro replaced it in February 2026. Gemini 3.1 Pro has stayed in preview status ever since, with no general-availability date announced as of this writing – it is currently the only Pro-tier Gemini 3 model callable through the API at all. Tracking gemini api costs per model over time shows the Flash line moving faster than the Pro line: three separate Flash versions have shipped on the same rate card since the Gemini 3.1 Pro preview release, while the Pro tier itself has not changed price.

Gemini API Pricing FAQ