Gemini API Pricing: Free Tier Limits and Paid Token Costs

Gemini API pricing has two layers: a free tier that lets you use Gemini Flash and Flash-Lite models at no cost within rate limits, and a paid tier billed per million tokens. As of September 2026, Gemini 3.8 Flash costs $0.75 per million input tokens and $3.75 per million output tokens, Gemini 3.1 Flash-Lite costs $0.25 and $1.50, and the preview Gemini 3.1 Pro costs $2 and $12 for prompts up to 200K tokens.

There is one date every Gemini developer needs in their calendar: on January 1, 2027, the 3.6, 3.7 and 3.8 Flash models double in price to $1.50 input and $7.50 output. This guide covers the free tier, every paid rate, that upcoming increase, caching storage fees, grounding costs and how to keep your bill low.

Gemini API pricing table: paid Standard tier

All prices below are USD per 1 million tokens, from Google’s official Gemini API pricing page. If you need a refresher on what a token is, read our plain English guide to tokens.

A few details matter here. Gemini 3.5 Flash-Lite charges the same $0.30 for text, image, video and audio input, while 3.1 Flash-Lite charges $0.25 for text, image and video but $0.50 for audio. Gemini 3.5 Flash also has a Priority tier at $2.70 input and $16.20 output for faster, more reliable processing.

The January 2027 Gemini Flash price increase

The current Flash prices come with an end date. Google’s pricing page shows that Gemini 3.8 Flash, 3.7 Flash and 3.6 Flash keep their $0.75 / $3.75 rates until December 31, 2026. From January 1, 2027, all three move to:

• Input: $1.50 per 1M tokens (up from $0.75)
• Output: $7.50 per 1M tokens (up from $3.75)
• Context caching: $0.15 per 1M tokens (up from $0.075)
• Cache storage: $1.00 per 1M tokens per hour (up from $0.50)

In other words, every Flash rate doubles. If you are building a budget or pricing your own product on top of Gemini Flash, use the 2027 numbers now so you are not caught out in three months.

Gemini API free tier: what you get and the limits

Google’s free tier is one of the best ways to start building with AI at no cost. According to the pricing page, the Flash and Flash-Lite models are “Free of charge” on the free tier. Gemini 3.1 Pro is not available on the free tier at all.

What Google no longer does is publish exact free tier rate limits in its docs. The rate limits documentation now tells you to check your limits inside Google AI Studio. A developer forum post from early September 2026 reported free limits of around 20 requests per day for Gemini 3.8 Flash and around 500 per day for Flash-Lite. That is a user report, not an official number, so check your own AI Studio dashboard.

If those figures are right, the free tier is fine for learning, prototypes and personal scripts, but not for a public app. Flash-Lite’s more generous allowance makes it the better choice for free experiments. For more zero-cost options, see our roundup of the best free AI tools.

Paid usage tiers and billing caps

Once you link a billing account, Google places your project in a usage tier. Higher tiers raise your rate limits and your billing cap. Qualification is based on how much you have paid and for how long.

The Tier 1 cap of $250 is a useful safety net for small projects. If you expect heavy traffic at launch, plan ahead: moving to Tier 2 takes both spend and at least three days.

Worked examples: what the Gemini API really costs

Here is the arithmetic on four typical workloads. Our guide to estimating AI API costs walks through the method in more detail.

Example 1: a chatbot with 1,000 conversations a day

Each conversation uses about 2,000 input tokens and 500 output tokens, so 2M input and 0.5M output tokens per day.

Example 2: the Gemini 3.1 Pro 200K threshold

A single 300,000-token prompt with a 5,000-token answer on 3.1 Pro crosses the 200K line, so it is billed at the higher rates: (0.3 x $4) + (0.005 x $18) = $1.20 + $0.09 = $1.29. If you could split it into two prompts under 200K, the same tokens at the lower rate would cost (0.3 x $2) + (0.005 x $12) = $0.60 + $0.06 = $0.66. Our context window explainer covers how to chunk long documents.

Example 3: context caching a large document

You load a 100,000-token product manual and ask 200 questions about it over 8 hours on 3.8 Flash.

• Without caching: 200 x 0.1M x $0.75 = $15.00 in input.
• With caching: reads cost 200 x 0.1M x $0.075 = $1.50, plus storage of 0.1M x $0.50 x 8 hours = $0.40, for about $1.90.

Unlike some rivals, Gemini charges an hourly storage fee for cached content, so caching pays off when a cache is read many times, and it can cost more than it saves if you store a big cache and barely use it. Our prompt caching guide compares how each provider handles this.

Example 4: a grounded research assistant

Your app makes 20,000 requests a month with Grounding with Google Search. The first 5,000 are free, and the other 15,000 cost 15 x $14 = $210, on top of token costs. Grounding is powerful but it can outweigh token spend for search-heavy apps.

Batch and Flex: half-price Gemini API

Google’s Batch and Flex options are both 50% off Standard rates. Gemini 3.8 Flash in batch mode costs $0.375 input and $1.875 output, and 3.5 Flash-Lite costs $0.15 and $1.25. Batch suits jobs that do not need an instant answer: bulk classification, nightly summaries, dataset labeling and content rewrites.

Example: summarizing 5,000 reports at 4,000 tokens in and 400 out is 20M input and 2M output tokens. On 3.8 Flash Standard: (20 x $0.75) + (2 x $3.75) = $15 + $7.50 = $22.50. In batch mode: $11.25. Read our Batch API guide for a cross-vendor comparison.

Gemini API vs OpenAI and Claude on price

Gemini’s biggest advantage is the free tier: no other major lab clearly offers free production-grade models for testing. On paid list price, OpenAI’s GPT-6 Luna undercuts Gemini’s budget models, and Gemini 3.1 Pro’s output rate is slightly above GPT-6.1 Sol and Claude Sonnet 5.5. See our OpenAI API pricing guide and Claude API pricing guide for full rate cards, or how to pick the cheapest model for each task.

Gemini API vs Google AI Pro subscription

The Gemini API and Google’s consumer subscriptions are separate products. Google AI Pro costs $19.99 a month in the US and Rs 1,950 a month in India, and gives you the Gemini app, Gemini in Gmail and Docs, 5 TB of storage and more. It does not include API usage. Developers pay for the API through Google AI Studio or Google Cloud billing. For the consumer plans, see our Gemini pricing guide, and for everyday office use, our guide to using Gemini in Gmail, Docs and Sheets.

Which Gemini model should you use?

• Students and hobbyists: the free tier with Flash-Lite, then Flash for harder prompts.
• Startups building chat or content features: Gemini 3.8 Flash, but budget at the 2027 price.
• High-volume classification or extraction: 3.1 Flash-Lite or 3.5 Flash-Lite, ideally through Batch.
• Complex reasoning and coding: Gemini 3.1 Pro, keeping prompts under 200K tokens where possible. Remember it is still a preview model, so test for stability.
• Research tools that need fresh facts: any Flash model with grounding, watching the $14 per 1,000 fee after 5,000 free requests.

Frequently asked questions

Our verdict on Gemini API pricing

The Gemini API is the easiest place to start for free, and its Flash-Lite models are among the cheapest ways to run bulk AI work. Gemini 3.8 Flash is excellent value right now, but its price doubles on January 1, 2027, so plan for $1.50 / $7.50 in any long-term budget. For heavy reasoning, 3.1 Pro is competitive if you stay under the 200K prompt threshold.

Use Batch for anything that can wait, cache only what you will reuse often, and keep an eye on grounding fees. Google changes rates and limits regularly, so confirm the current numbers on the official Gemini API pricing page before you commit.