Gemini API Pricing 2026: Costs, Free Tier & Models

The Gemini API uses usage-based pricing rather than a simple monthly subscription. Google offers a Free Tier for eligible models, while paid usage is generally charged per million tokens, with different prices for input, output, caching, and some tools. Current Gemini models include options such as Gemini 3.5 Flash, Gemini 3.1 Flash-Lite, Gemini 3.1 Pro Preview, Gemini 2.5 Flash, and Gemini 2.5 Pro.

Gemini API Pricing at a Glance

If you are trying to understand how much the Gemini API costs, the most important thing is that there is no single Gemini API price.

Your cost depends on:

• The Gemini model you use
• Number of input tokens
• Number of output tokens
• Whether prompts exceed certain context thresholds
• Cached input
• Audio or other modalities
• Google Search or Maps grounding
• Standard, Batch, Flex, or Priority processing
• How frequently your application makes requests

Google's current pricing page lists separate Free and Paid tiers, with additional enterprise options for larger deployments.

Current Gemini API Pricing

Google prices Gemini API usage primarily per 1 million tokens (1M tokens).

The following table focuses on major current text-capable models and their standard paid pricing.

*For the Pro models shown above, higher pricing applies when prompts exceed 200K tokens.

Which Gemini API model is cheapest?

Among the major current models listed by Google, Gemini 2.5 Flash-Lite has the lowest standard token price at $0.10 per 1M input tokens and $0.40 per 1M output tokens.

Gemini 3.1 Flash-Lite is the lower-cost option in the newer Gemini 3 family at $0.25 per 1M input tokens and $1.50 per 1M output tokens.

However, the cheapest model is not automatically the best model. For example:

• A simple classification task may work well with Flash-Lite.
• A normal AI application may benefit from Gemini 2.5 Flash.
• Complex coding or reasoning may justify a Pro model.
• Advanced agentic or coding workloads may benefit from Gemini 3.5 Flash. Google describes Gemini 3.5 Flash as its most intelligent Flash model for sustained agentic and coding workloads.

Interactive Gemini API Cost Estimator

Use our interactive calculator to compare model costs, calculate token usage, and see how caching or Batch discounts reduce your monthly bill.

What Is the Gemini API?

The Gemini API lets developers integrate Google's Gemini models into their own software, websites, applications, automation systems, and AI workflows.

Instead of opening the Gemini consumer app and manually asking questions, developers can send requests programmatically.

For example, an application could use the API to:

1. Receive a customer's question.
2. Send the question to a Gemini model.
3. Receive the model's response.
4. Display the answer inside the application.

Google provides official SDKs and documentation for getting started. Developers can create an API key through Google AI Studio and use the Gemini API directly.

Is Gemini API Free?

Yes, but only within Google's Free Tier limits and for eligible models.

Google says new accounts begin on the Free Tier, which provides access to certain Gemini API models and AI Studio subject to model-specific rate limits.

The Free Tier can be useful for:

• Learning the Gemini API
• Building prototypes
• Testing prompts
• Creating small projects
• Experimenting with different models
• Developing proof-of-concept applications

But free access does not mean unlimited API usage. Your application can encounter rate limits, model restrictions, or other limits depending on the model and current Google policies.

Google also states that content on the Free Tier may be used to improve its products, while the paid tier provides different data-use terms.

Gemini API Free Tier vs Paid Tier

Google's current pricing documentation specifically distinguishes Free, Paid, and Enterprise offerings.

How Does Gemini API Token Pricing Work?

This is the part that confuses many first-time API users. Gemini does not normally charge you a fixed amount for every API request.

Instead, billing is based largely on tokens. There are two basic categories:

Input tokens

These are the tokens you send to Gemini. They can include your prompt, instructions, conversation history, text documents, or other supported input data.

Output tokens

These are the tokens generated by Gemini in its response. For models with thinking enabled, Google's pricing documentation can include thinking tokens in the output price.

Gemini API Cost Formula

A simple way to estimate your API cost is:

For example, suppose an application sends 100,000 input tokens and receives 20,000 output tokens. Using a model priced at $0.30 per 1M input tokens and $2.50 per 1M output tokens, the calculation is:

• Input: 100,000 ÷ 1,000,000 × $0.30 = $0.03
• Output: 20,000 ÷ 1,000,000 × $2.50 = $0.05
• Total request cost: $0.03 + $0.05 = $0.08

This is why looking only at the price per API request can be misleading. Your actual bill depends on how much data your application sends and how much the model generates.

Why Output Tokens Can Make Gemini API More Expensive

One important pricing pattern is that output tokens are often considerably more expensive than input tokens.

For example, Gemini 2.5 Flash is currently listed at:

• $0.30 per 1M input tokens
• $2.50 per 1M output tokens

That means generating large responses can cost much more than sending the same token volume as input. This matters for applications that generate long articles, large code files, detailed reports, long conversations, large JSON responses, or extensive reasoning outputs.

Gemini API Pricing for Long Prompts

Some Gemini Pro models have different prices depending on prompt size. For example, Google's current Gemini 2.5 Pro pricing is:

Gemini 3.1 Pro Preview similarly has higher pricing above 200K tokens:

This is important if you are building applications that process large books, large codebases, extensive conversation histories, huge datasets, or long RAG contexts.

What Is Gemini API Context Caching?

Context caching can reduce the cost of repeatedly sending the same information to Gemini.

Imagine you have a large document that your application sends with hundreds of requests. Instead of treating the entire document as fresh input every time, caching can reduce the cost of repeated context.

Google says implicit caching is enabled by default for Gemini 2.5 and newer models, while explicit caching is available through the appropriate API workflow. Caching can be useful for applications involving long documents, repeated conversations, large system instructions, RAG applications, or long-running workflows.

However, caching has its own pricing rules, including cache storage charges on applicable models.

Does Gemini API Have a Batch Discount?

Yes. Google's Gemini Batch API is designed for asynchronous workloads and is priced at 50% of the standard cost for supported requests. Google gives examples such as data preprocessing and evaluations where an immediate response is not necessary.

This can make Batch attractive for large-scale data processing, dataset analysis, offline classification, evaluations, bulk content processing, or background jobs.

For example, if a workload would normally cost $20 using standard pricing, a supported Batch workload would cost approximately $10 at the 50% standard-cost rate. The trade-off is that Batch is asynchronous rather than designed for immediate interactive responses.

Gemini API Standard vs Batch vs Flex vs Priority

Google also offers different processing options for some models.

Pricing varies by model and processing type, so developers should check the model-specific pricing rather than assuming one discount applies everywhere.

Does Gemini API Charge for Google Search Grounding?

It can. Some Gemini models support grounding with Google Search or Google Maps.

Google's pricing documentation lists separate charges for grounding after applicable free allowances. For example, the current Gemini 3 pricing documentation lists 5,000 free Google Search grounding requests per month shared across Gemini 3 models, followed by additional per-query charges.

This means your application cost can be higher than simply token calculations. You also need to consider additional tools and services your application invokes.

Gemini API Billing Architecture

To understand what determines your monthly API bill, it helps to look at the billing flow:

How Much Does 1,000 Gemini API Requests Cost?

There is no universal answer. 1,000 API requests do not automatically equal a fixed price.

Consider two different applications making 1,000 requests:

• App A (Lightweight Chat): Each request sends 500 input tokens and receives 200 output tokens. Under Gemini 2.5 Flash, 1,000 requests cost **$0.65**.
• App B (Heavy Document RAG): Each request sends 50,000 input tokens and receives 10,000 output tokens. Under Gemini 2.5 Flash, 1,000 requests cost **$40.00**.

Both make exactly 1,000 requests, but App B costs 60 times more due to the large token volume.

Gemini API Cost Example

Suppose your application uses Gemini 2.5 Flash and processes 10 million input tokens and 2 million output tokens over the month.

At the standard rates ($0.30/1M input, $2.50/1M output), the math looks like this:

• Input Cost: 10 × $0.30 = $3.00
• Output Cost: 2 × $2.50 = $5.00
• Estimated Monthly Total: $3.00 + $5.00 = $8.00

This example assumes no grounding tools, cache storage charges, or other services.

Which Gemini API Model Should You Choose?

Google provides different models targeting specific performance/cost tradeoffs:

• Gemini 2.5 Flash-Lite (Cheapest): Stated standard price is $0.10 input / $0.40 output. Excellent for high-volume classifications, data extractions, and lightweight automations.
• Gemini 3.1 Flash-Lite (Cheapest in Gemini 3): Priced at $0.25 input / $1.50 output. Ideal for speed, scale, and low-cost newer generation tasks.
• Gemini 2.5 Flash (Balanced): Priced at $0.30 input / $2.50 output. A great general-purpose production option that balances capability, speed, and cost.
• Gemini 2.5 Pro (Reasoning & Coding): Priced at $1.25 input / $10.00 output. State-of-the-art model that excels at complex reasoning and developer coding tasks.
• Gemini 3.5 Flash (Frontier Agents): Priced at $1.50 input / $9.00 output. Optimized for sustained agentic execution, heavy coding, and long-horizon tasks.

Gemini API vs Gemini Subscription: Are They the Same?

No. A consumer Gemini subscription (like Gemini Advanced) and Gemini API usage are separate concepts.

If you are building an application with an API key, you need to configure Gemini API billing through Google AI Studio or Google Cloud. A consumer subscription does not cover API usage.

• Gemini App Subscription: Flat monthly fee designed for using Google's Gemini consumer application directly.
• Gemini API: Usage-based token charging designed for developers integrating Gemini into custom software.

How to Start Using the Gemini API

1. Open Google AI Studio: Log in and create a project in the Google AI Studio console.
2. Create an API Key: Generate an API key through the AI Studio creation workflow.
3. Install the SDK: Install Google's supported client library. For example, for Python:

   pip install -U google-genai

4. Initialize the Client: Pass your API key to client SDK libraries to make your first calls.
5. Upgrade to Paid Tier: To increase rate limits for production, connect a Google Cloud billing account. Upgrading requires a minimum prepayment of $10 in credit.

How to Reduce Gemini API Costs

Better API architecture can significantly reduce developer spending:

• Choose the Cheapest Model: Match your task difficulty to the model. Don't use Gemini Pro for simple text sorting.
• Limit Output Length: Output tokens are much more expensive than input tokens. Set strict `max_output_tokens` rules in your configuration.
• Use Context Caching: Cache large repeating contexts (e.g. system prompts, large RAG documents) to reduce input cost.
• Leverage Batch API: Run non-urgent offline workloads through the Batch API to save 50% on token rates.
• Prune Chat History: Avoid sending unnecessary historic messages. Keep only the conversation context required for immediate reasoning.

Is Gemini API Good for Production?

Yes, through Google's Paid Tier. Google offers production-grade paid API access with higher rate limits, enterprise availability, and data privacy safeguards. Under the Paid Tier, Google states that content is not used to train its AI models.

Developers should evaluate peak traffic, token budgets, rate limits, and model lifecycle deprecation policies. Google maintains a dedicated model deprecation schedule, noting that older models are retired over time. For example, Gemini 2.0 Flash and Flash-Lite were shut down on June 1, 2026, and Gemini 3.1 Flash-Lite has an announced shutdown of May 7, 2027.

Frequently Asked Questions

Final Verdict

Gemini API pricing is highly competitive for developers who choose the right model and control token usage.

The best developer strategy is not to automatically choose the cheapest model, but to select the least expensive model that reliably completes the job. This approach ensures the best balance of API cost, output quality, execution speed, and application scalability.