Gemini API Price Increase: Dec 1 and Jan 1

The Gemini API price increase is already written into Google's own pricing page, and it comes in two steps. Gemini 4 Argon gets twice as expensive on December 1, 2026. The Gemini 3.8 Flash and 3.6 Flash models follow on January 1, 2027, with the same doubling. If you build anything on these models, the dates matter more than any launch announcement, because they quietly change your monthly bill. You can test your own numbers with our AI API cost calculator, which already knows about both dates.

This guide is about money, not hype. We read the official Gemini API pricing page and wrote down exactly which models change, which tiers change, and what the new rates are. Then we turned the percentages into real monthly examples, because "price doubles" sounds very different when it is $15 and when it is $15,000. Wherever something is our own arithmetic, we say so, and wherever we could not confirm a detail, we flag it.

The Gemini API price increase in two dates

Google's pricing page does not hide the change. Next to each affected rate, it states an end date for the lower price and a start date for the higher one. Here is what is listed:

• December 1, 2026: Gemini 4 Argon moves to its higher price. This covers the Standard, Batch, Flex, Priority and Contributor tiers. Priority, for example, goes from $3.60 for input and $18 for output to $7.20 and $36.
• January 1, 2027: Gemini 3.8 Flash changes on the Standard, Batch, Flex, Priority and Contributor tiers. Gemini 3.6 Flash changes on Standard, Batch, Flex and Priority.
• January 1, 2027, smaller models: Gemini 3.8 Flash TTS and Gemini 3.8 Flash-Lite TTS, plus two Gemini Robotics ER 2 preview models, also have dated changes.

Every other model on the page shows no dated price change. One more date is worth writing down even though it is not a price: the page lists a shutdown date of March 15, 2027 for Gemini 2.5 Flash Image.

If you are new to the model itself, our comparison of Gemini 4 Argon against GPT-6 and Muse Spark Open explains what it is good at. This article only deals with what it costs.

Gemini 4 Argon price increase on December 1

Gemini 4 Argon is the first to change. For the Standard paid tier, per million tokens, Google's page lists:

• Input: $2.00 through November 30, then $4.00 from December 1.
• Output, including thinking tokens: $10.00 through November 30, then $20.00 from December 1.
• Context caching: $0.10 through November 30, then $0.20 from December 1. The page notes that only implicit caching is available for this model.

Gemini 4 Argon is not available on the free tier, so there is no free path around the change. If you are paying for it today, you are paying the lower rate for about a month more.

A token, in case the unit is unfamiliar, is a small chunk of text, roughly three quarters of a word in English. Models read and write in tokens, and pricing is quoted per million of them. If you want the basics of what the model is actually doing with those tokens, our beginner explainer on how AI works covers it without jargon.

Gemini 3.8 Flash price in January 2027 (and 3.6 Flash)

The Flash models are the workhorses many apps use for chat, summaries and classification, the everyday uses of AI beyond chatbots, and they have the biggest set of tiers. Search coverage often calls the lower rates "introductory pricing," and Google's cloud pricing does use that word. The Gemini API page itself is plainer: it simply shows each rate with the date it ends. The page shows the same figures for Gemini 3.8 Flash and Gemini 3.6 Flash. Per million tokens, the lower price runs through December 31, 2026 and the higher one starts January 1, 2027:

Standard

• Input: $0.75, then $1.50
• Output, including thinking: $3.75, then $7.50
• Context caching: $0.075, then $0.15
• Cache storage, per million tokens per hour: $0.50, then $1.00

Batch and Flex

• Input: $0.375, then $0.75
• Output: $1.875, then $3.75
• Context caching: $0.0375, then $0.075
• Cache storage: $0.50, then $1.00

Priority

• Input: $1.35, then $2.70
• Output: $6.75, then $13.50
• Context caching: $0.135, then $0.27
• Cache storage: $0.50, then $1.00

So every line doubles. The free tier is the exception: the page lists input, output and caching as free for both Flash models, with the note that free-tier content is used to improve Google's products. The page also describes Gemini 3.8 Flash as Google's most intelligent Flash model and 3.6 Flash as the previous generation.

One caution if you use Google's cloud platform instead of the Gemini API. The Vertex AI pricing is a separate page, and a report of it lists Gemini 3.7 Flash next to 3.6 Flash with the same dates and the same $0.75 and $3.75 starting rates. We could not open that page in full, so if you build on Vertex, read its table yourself before you plan. One third-party source also says that on Vertex AI the lower Argon rate is delivered as credits back on net spend, which would change your effective cost there. We could not confirm that, so treat it as a reason to check, not as a fact.

How to read the pricing page without getting caught out

Before the examples, a few habits that save money on any AI pricing page, not only Google's.

• Check the date next to the price. A rate with an end date is a rate that will change. Copying today's number into a budget for next year is the most common mistake.
• Check which tier you use. Standard, Batch, Flex and Priority have different prices, and the increase applies to each of them. Your cost depends on the tier your code actually calls.
• Remember that thinking counts as output. The page says the output price includes thinking tokens, so a model that reasons for a long time costs more than the length of its final answer suggests.
• Look for extras. Search grounding, for instance, is billed separately: the page describes 5,000 free search requests a month shared across Gemini 3 and newer models, then $14 per 1,000 requests, with a daily cap of 1,500 requests.
• Check the platform. The Gemini API and Google's cloud platform have separate pages with separate model lists.

If you are planning Gemini API pricing for 2027, use the January rates in your budget from the start. It is much easier to be pleasantly surprised than to explain a doubled invoice to your team.

What the increase costs in real money

Percentages hide the effect, so here are examples. These are our own arithmetic from the page's prices, assuming a month of standard-tier use with no caching. Your usage will differ, so treat them as shapes, not forecasts.

A small app on Gemini 3.8 Flash: 10 million input tokens and 2 million output tokens a month.

• Until December 31: 10 × $0.75 plus 2 × $3.75 comes to $7.50 plus $7.50, which is $15.00.
• From January 1: 10 × $1.50 plus 2 × $7.50 comes to $15.00 plus $15.00, which is $30.00.

A busier product: 200 million input and 40 million output tokens a month.

• Before: $150 plus $150, which is $300.
• After: $300 plus $300, which is $600.

The same small app on Gemini 4 Argon: 10 million input and 2 million output.

• Until November 30: 10 × $2 plus 2 × $10 is $20 plus $20, which is $40.
• From December 1: 10 × $4 plus 2 × $20 is $40 plus $40, which is $80.

The pattern is simple. Because both input and output prices double, your bill doubles no matter what the mix is. It only stays flat if your usage drops by half, or if you move work to something cheaper.

Why a flat doubling is more serious than it looks

For a hobby project, an extra $15 a month is a nuisance. For a product with thin margins, it is not. If your pricing to customers is fixed and AI is your largest cost, a doubled model price comes straight out of your profit unless you act.

There is also a timing trap. The increase lands on January 1, right when many teams reset budgets and often freeze changes for the holidays. If you leave the work until the new year, you will pay the higher price for at least a few weeks while you decide. That is the same lesson developers learned this month when a hosting platform announced an end date and people searched is Deno dead: when a vendor sets the rules, the safest move is to plan before the date, not after it.

How to prepare, in the order we would do it

1. Find out what you actually spend. Look at your last full month of usage, split by model and by feature. Many teams discover that one feature, often a long-context summary or an agent loop, causes most of the bill.
2. Run the numbers for both dates. Put your monthly input and output tokens into the cost calculator. It shows today's price and the scheduled one side by side, so you can see the real jump for your own usage.
3. Use caching where it fits. Cached input is far cheaper than fresh input, even after the increase. If your prompts repeat a long instruction or document, caching is the biggest lever you have. Remember that cache storage also doubles, so check the hours your cache stays alive.
4. Move non-urgent work to Batch or Flex. On the page, Batch and Flex cost half the Standard rate. After the increase, a Batch job on Gemini 3.8 Flash costs the same as a Standard job does today. Anything that can wait, such as nightly summaries or bulk classification, belongs there.
5. Shorten what you send. Output tokens cost five times as much as input on these models, so a model that rambles is expensive. Ask for short answers and limit reasoning where you do not need it.
6. Test a cheaper model on your real task. Our calculator lists several. For example, as of our last check, Gemini 2.5 Flash-Lite is $0.10 for input and $0.40 for output, and Gemini 3.5 Flash-Lite is $0.30 and $2.50. On the same 10 million and 2 million token example, that is about $1.80 and $8.00 a month. These models are not equal in ability, so measure quality before you switch, not just price.
7. Set a budget alert. Whatever you decide, put a spending limit and an alert in place before December 1, so a surprise shows up in a notification and not an invoice.

Most of these steps are just habits. As we argue in our piece on why most people use AI wrong, the people who get the best results from AI treat it as a tool with a cost and a purpose, not as magic that you leave running.

Should you switch to another provider?

Maybe, but do it for a reason, not out of panic. A model that is cheaper on paper can cost more once you count the extra prompts, retries and engineering time. Compare on your own task, with your own data, and include the output tokens, since that is where the price difference is often biggest.

If you do compare, use the same unit everywhere. The calculator turns every provider into a monthly cost for your input and output volume, which makes a fair comparison much easier than reading five pricing pages. Keep in mind that other providers also change their prices, tiers and models on short notice, so the cheapest option today may not stay that way.

Our advice is to keep your code flexible. If the model name and the provider live in one configuration file, switching later is a small job. If they are scattered through your app, every price change becomes a project.

What we still do not know

• Whether Google will extend the lower prices or announce further changes before the dates. Pricing pages can be updated at any time, so check the live page before you budget.
• How the Vertex AI page matches the Gemini API page, since we could not read it in full and models are named differently.
• Whether other tiers or smaller models will get dated changes later. At the time of writing, the page lists none for them.
• Whether newer Flash models will arrive at lower prices that make the old ones less attractive.

We will update this article when any of these change. The safest assumption is that the dates on the page are real, and that the rates on the page are the ones your bill will follow.

FAQs

Is the Gemini API price increase confirmed?

Yes. Google's Gemini API pricing page lists dated prices: Gemini 4 Argon is $2 for input and $10 for output per million tokens through November 30, 2026, and $4 and $20 from December 1. Gemini 3.8 Flash and 3.6 Flash are $0.75 and $3.75 through December 31, and $1.50 and $7.50 from January 1, 2027.

Which Gemini models are getting more expensive?

Gemini 4 Argon on December 1, 2026. Gemini 3.8 Flash and Gemini 3.6 Flash on January 1, 2027. Gemini 3.8 Flash TTS, Gemini 3.8 Flash-Lite TTS and two Gemini Robotics ER 2 preview models also have January 1 changes. Other models on the page show no dated change.

Does the price increase apply to the free tier?

For the Flash models, the page lists input, output and context caching as free on the free tier. Gemini 4 Argon is not available on the free tier. Free-tier content is used to improve Google's products, so check that before sending anything private.

How much will my bill go up?

Both input and output prices double, so your bill doubles for the same usage. A small app using 10 million input and 2 million output tokens a month on Gemini 3.8 Flash goes from about $15 to about $30. Use the AI API cost calculator for your own volume.

Can Batch or Flex pricing soften the increase?

Yes, partly. On the page, Batch and Flex cost half the Standard rate for these models. After the increase, Batch pricing on Gemini 3.8 Flash equals today's Standard price. It only suits work that can wait, not live chat.

Is the Vertex AI price the same as the Gemini API price?

Not necessarily. Vertex AI has its own pricing page, and a report of it lists Gemini 3.7 Flash alongside 3.6 Flash with the same dates. Check the page for the platform you actually use.

What should I do before December 1?

Measure your current spend by model and feature, test caching and batch processing, try a cheaper model on your real task, and set a budget alert. If you use Gemini 4 Argon, the first change comes on December 1.