Is the Gemini API Getting More Expensive? What Changes on 1 December
By Ashikur Rahman13 min readQuick answer: Yes. Google's own Gemini API pricing page shows three changes. On 1 December 2026, Gemini 4 Argon doubles: input goes from $2.00 to $4.00 per million tokens and output from $10.00 to $20.00. On the same day, requests to the deprecated Gemini 2.5 Pro and 2.5 Flash start routing automatically to Gemini 3.8 Flash and 3.5 Flash-Lite and are billed at those rates. Then on 1 January 2027, Gemini 3.8 Flash and 3.6 Flash double from $0.75 to $1.50 input and $3.75 to $7.50 output. Gemini 3.5 Flash-Lite shows no scheduled change. If you call Argon at any volume, your bill for the same work doubles in three weeks.
What is changing in Gemini API prices, and when?
Everything below comes from Google's Gemini API pricing page, read on 10 November 2026. The page now lists two prices for several models, one "through" a date and one "starting" the next day. Here is the full schedule for the paid tier, Standard processing, per million tokens in US dollars.
| Model | Price today | New price | From |
|---|---|---|---|
| Gemini 4 Argon, input | $2.00 | $4.00 | 1 December 2026 |
| Gemini 4 Argon, output (incl. thinking) | $10.00 | $20.00 | 1 December 2026 |
| Gemini 4 Argon, context caching | $0.10 | $0.20 | 1 December 2026 |
| Gemini 3.8 Flash and 3.6 Flash, input | $0.75 | $1.50 | 1 January 2027 |
| Gemini 3.8 Flash and 3.6 Flash, output | $3.75 | $7.50 | 1 January 2027 |
| Gemini 3.8 Flash and 3.6 Flash, caching | $0.075 | $0.15 | 1 January 2027 |
| Flash cache storage, per million tokens per hour | $0.50 | $1.00 | 1 January 2027 |
| Gemini 3.5 Flash-Lite | $0.30 input, $2.50 output | No change shown | None |
| Gemini 2.5 Pro | $1.25 input, $10.00 output (prompts up to 200k) | Routed to 3.8 Flash, billed at 3.8 Flash rates | 1 December 2026 |
| Gemini 2.5 Flash | $0.30 input, $2.50 output | Routed to 3.5 Flash-Lite, billed at Flash-Lite rates | 1 December 2026 |
Every change is a doubling or a reroute, and none is a cut except for 2.5 Pro users, whose rerouted requests get cheaper for a month before the Flash increase lands. Google's page does not give a reason for any of the increases.
The Argon price that ends on 30 November was its launch price. Google announced Gemini 4 Argon on 13 October 2026, and coverage at the time reported it at $2 per million input tokens and $10 per million output tokens. The pricing page now shows that figure ending after roughly seven weeks.
What does Gemini 4 Argon cost after 1 December?
Double, on every line. Google's page lists Argon at "$2.00 through November 30, 2026. $4.00 starting December 1, 2026" for input, and "$10.00 through November 30, 2026. $20.00 starting December 1, 2026" for output, which includes thinking tokens. Context caching moves from $0.10 to $0.20.
Three Argon details make the increase bite harder than the headline suggests:
- No free tier. The page says "Gemini 4 Argon is not available on the Free tier." There is no free allowance to absorb testing.
- Thinking tokens are output. "Every token the model processes is counted and billed, including thinking tokens and internal tool system instructions and function declarations when using grounding." Thinking is billed at the output rate, which is the line going to $20.00.
- Implicit caching only. Argon supports "Implicit caching only (explicit caching is not supported)." You cannot pin a large shared prompt in the cache yourself to cut repeat costs the way you can on some other models.
Argon is Google's flagship. The page describes it as "built for complex software engineering, deep mathematical and scientific problem solving, autonomous multi-step agents, and long-decode generation up to 1M tokens." Long outputs and multi-step agents are exactly the workloads that produce a lot of output and thinking tokens, so they feel the $20.00 line most.
Why are Gemini 2.5 Pro and 2.5 Flash being rerouted?
Because Google has deprecated them. The page says Gemini 2.5 Pro "is deprecated and is no longer available to select in Google AI Studio. Effective December 1, 2026, all API requests to gemini-2.5-pro will automatically route to Gemini 3.8 Flash (gemini-3.8-flash) and be billed at Gemini 3.8 Flash rates." Gemini 2.5 Flash gets the same treatment, routing to Gemini 3.5 Flash-Lite "and be billed at Gemini 3.5 Flash-Lite rates."
This matters even if you do nothing, perhaps especially then. Your code keeps calling gemini-2.5-pro, the call keeps succeeding, and a different model answers. Two things change silently:
- The model. A request written and tested against a Pro model is answered by a Flash model. Output quality, length, format and reasoning depth can all differ. Prompts tuned for 2.5 Pro may need retuning.
- The price. 2.5 Pro was listed at $1.25 input and $10.00 output for prompts up to 200k tokens. Routed to 3.8 Flash, the same request costs $0.75 and $3.75 in December, then $1.50 and $7.50 from 1 January 2027. 2.5 Flash, at $0.30 and $2.50, routes to 3.5 Flash-Lite at the same $0.30 and $2.50.
Gemini 2.5 Flash-Lite, at $0.10 input and $0.40 output, carried no deprecation notice on the page when it was read. The page does not say whether it will be rerouted later.
What does a typical workload cost before and after?
Here is one sample workload priced on each model: 10 million input tokens and 2 million output tokens a month, Standard processing, no caching, no grounding. Your mix will differ, but the ratios hold.
| Model | November 2026 | December 2026 | From January 2027 |
|---|---|---|---|
| Gemini 4 Argon | $40.00 | $80.00 | $80.00 |
| Gemini 3.8 Flash | $15.00 | $15.00 | $30.00 |
| Gemini 2.5 Pro (routed to 3.8 Flash from 1 December) | $32.50 | $15.00 | $30.00 |
| Gemini 2.5 Flash (routed to 3.5 Flash-Lite from 1 December) | $8.00 | $8.00 | $8.00 |
The arithmetic, so you can redo it with your own numbers. Argon today: 10 x $2.00 + 2 x $10.00 = $40.00. From 1 December: 10 x $4.00 + 2 x $20.00 = $80.00. Gemini 3.8 Flash today: 10 x $0.75 + 2 x $3.75 = $15.00, rising to $30.00 in January. Gemini 2.5 Pro at its listed rate: 10 x $1.25 + 2 x $10.00 = $32.50, which falls to the 3.8 Flash figure once routing starts. Gemini 2.5 Flash: 10 x $0.30 + 2 x $2.50 = $8.00, unchanged after routing because 3.5 Flash-Lite has the same rates.
After both increases, Argon still costs about 2.67 times as much as 3.8 Flash per token, on input ($4.00 against $1.50) and output ($20.00 against $7.50) alike. The gap between the flagship and the workhorse stays the same; both just cost twice as much as they did in November.
The Batch option
Google's paid-tier summary lists "Batch API (50% cost reduction)". If that discount applies to Argon after 1 December, the sample workload would come to $40.00 a month on Batch, the same as today's Standard price. The Batch tab for Argon was not read for this article, so confirm the post-December Batch rate on the pricing page before you plan around it. Batch suits work that does not need an answer in seconds: overnight document processing, bulk classification, evaluation runs.
Transparency note: CriticNest has no affiliate or commercial relationship with Google, Cursor or Windsurf, and no link in this article earns a commission. Every price and quotation was read on Google's Gemini API pricing page on 10 November 2026, and the cost examples are arithmetic on those prices, not measured bills. No model was benchmarked for this article.
How to estimate your own December bill
You do not need a calculator tool to see what 1 December does to you. You need three numbers from your last full month of usage, which your Google Cloud billing reports or AI Studio usage page will show by model.
- Input tokens per model, in millions. Write down how many million input tokens you sent to each model last month.
- Output tokens per model, in millions. Do the same for output, remembering that output includes thinking tokens on Argon and the Flash models. If your report splits thinking out, add it back in.
- Grounded searches. Note how many Google Search requests your grounded calls triggered, and subtract the 5,000 free.
Then apply the new rates. For Argon, multiply input millions by $4.00 and output millions by $20.00. For anything still calling gemini-2.5-pro, use the 3.8 Flash rates of $0.75 and $3.75 for December, and $1.50 and $7.50 from January. For 3.8 Flash and 3.6 Flash, keep today's rates for December and double them for January. Add grounding at $14 per 1,000 searches beyond the free allowance.
A shortcut works for most projects: if Argon is your main model, your December token bill will be very close to twice November's, because every Argon line doubles at once. If Flash is your main model, December looks like November and January looks like twice November. If you mainly use 3.5 Flash-Lite, the current page gives you no reason to expect a change at all. The project that needs the most care is the one that mixes Argon with deprecated 2.5 Pro calls, because one part of the bill doubles while the other drops and changes model.
Costs that are easy to miss
Grounding with Google Search. For Gemini 3 and newer models, the page lists "5,000 free search requests per month (shared across all Gemini 3 and newer models), then $14 per 1,000 requests (subject to a daily cap of 1,500 requests per day)." It also warns that "A customer-submitted request to Gemini may result in one or more queries to Google Search. You will be charged for each individual search query performed." One user prompt can mean several billed searches. At the daily cap, a month is about 45,000 searches; minus the 5,000 free, that is 40,000 paid, or $560 a month in grounding alone. That figure does not change on 1 December, but it sits on top of the token bill.
Thinking tokens. On Argon and the Flash models, output prices include thinking tokens. A model that reasons at length before answering can spend more on thinking than on the visible answer, and from 1 December every one of those tokens on Argon costs $20.00 per million.
Cache storage. On the Flash models, explicit caching has two charges: a per-token caching price and a storage price per million tokens per hour. Both double on 1 January 2027, the storage line from $0.50 to $1.00. A large cache left running all month is not free.
The free tier's trade. The Flash models still have a free tier, but the page answers Yes on its "Used to improve our products" row, against No on the paid tier. If you are tempted to push production traffic onto the free tier to dodge the increase, your data goes into Google's product improvement in exchange.
How to cut your Gemini bill before the increase
- Find out what you actually call. Pull a month of usage by model. Many projects call Argon for tasks Flash would handle, simply because Argon was the default in a template.
- Route by task. Send hard reasoning and long code to Argon and everything else to 3.8 Flash or 3.5 Flash-Lite. Flash-Lite has no scheduled increase on the page, which makes it the stable choice for high-volume simple work: classification, extraction, translation, short summaries.
- Move batchable work to Batch. Anything that can wait should not pay Standard rates.
- Control thinking. Where your model and SDK let you limit reasoning, set it to what the task needs. Thinking billed at the output rate is the fastest-growing line on most agent bills.
- Stop calling deprecated names. Replace gemini-2.5-pro and gemini-2.5-flash in your code with the model you actually want, then test it. Do not let the 1 December routing choose for you.
- Watch grounding counts. Log how many searches each request triggers. If it is more than one on average, the free 5,000 goes quickly.
- Set budget alerts. Put a spending alert on your Google Cloud billing account at your November level, so December's doubling shows up as an alert rather than an invoice.
If you pay for AI through a tool, not the API
Many developers never touch the Gemini API directly; they pay a monthly subscription for an AI coding editor and let it choose the models. If that is you, these API prices reach you only indirectly, through whatever the tool decides to charge. Our reviews of whether Cursor is worth paying for and whether Windsurf is better than Cursor cover that side of the bill.
Does this affect Gemini app subscribers?
No. The API prices above apply to developers calling Google's models from code. People who use the Gemini app through a Google AI subscription pay a flat monthly fee and are not billed per token. They have their own change to deal with: from 13 November 2026, Google is removing Pro model selection from its $4.99 Google AI Plus plan. Our guide to whether Google AI Plus is still worth it covers that.
The two moves are separate, but they point the same way. On both the consumer and developer side, Google's cheapest routes to its strongest models are getting narrower or more expensive in the last weeks of 2026.
Who should rethink their Gemini setup now?
- Heavy Argon users. A doubling in three weeks is the largest change on the page. Audit what really needs Argon before 1 December.
- Anyone still on 2.5 Pro. Your bill may fall in December, but your model changes to a Flash model without a code change. Test now.
- Flash-heavy apps on thin margins. You have until 1 January 2027. If your pricing to customers assumes $0.75 input, it needs revisiting.
- Agent builders using search grounding. Your grounding cost does not change, but your token cost does, and agents tend to be heavy on both thinking and searching.
- Light users on Flash-Lite. Nothing changes for you on the current page.
The short version
Google's Gemini API pricing page shows Gemini 4 Argon doubling on 1 December 2026, from $2.00 to $4.00 per million input tokens and $10.00 to $20.00 output, with caching doubling too. The same day, requests to deprecated Gemini 2.5 Pro and 2.5 Flash route to 3.8 Flash and 3.5 Flash-Lite and are billed at those rates. On 1 January 2027, 3.8 Flash and 3.6 Flash double to $1.50 input and $7.50 output. Flash-Lite shows no change. On a sample 10 million input and 2 million output tokens a month, Argon goes from $40.00 to $80.00 and 3.8 Flash from $15.00 to $30.00. Route each task to the cheapest model that handles it, move batchable work to Batch, limit thinking, and replace deprecated model names yourself before Google's routing does it for you.
Frequently asked questions
Is the Gemini API getting more expensive?
Yes. Google's pricing page shows Gemini 4 Argon doubling on 1 December 2026, from $2.00 to $4.00 per million input tokens and from $10.00 to $20.00 per million output tokens. Gemini 3.8 Flash and 3.6 Flash double on 1 January 2027. Gemini 3.5 Flash-Lite shows no scheduled change.
When does the Gemini 4 Argon price increase start?
On 1 December 2026. The pricing page lists Argon's current rates "through November 30, 2026" and the new rates "starting December 1, 2026." The $2 and $10 rates were Argon's launch price when Google announced the model on 13 October 2026.
What happens to Gemini 2.5 Pro on 1 December 2026?
All API requests to gemini-2.5-pro automatically route to Gemini 3.8 Flash and are billed at 3.8 Flash rates. Google says 2.5 Pro is deprecated and no longer selectable in Google AI Studio. Your code keeps working, but a different model answers, so test your prompts before then.
What happens to Gemini 2.5 Flash?
From 1 December 2026, requests to gemini-2.5-flash route automatically to Gemini 3.5 Flash-Lite and are billed at Flash-Lite rates. Both are listed at $0.30 per million input tokens and $2.50 output, so the price stays the same, but the model answering changes.
How much will Gemini 3.8 Flash cost in 2027?
From 1 January 2027, $1.50 per million input tokens and $7.50 per million output tokens, double today's $0.75 and $3.75. Context caching goes from $0.075 to $0.15 and cache storage from $0.50 to $1.00 per million tokens per hour. Gemini 3.6 Flash has the same schedule.
Is Gemini 4 Argon free to try?
No. Google's pricing page says Gemini 4 Argon is not available on the Free tier, so every token is billed. The Flash and Flash-Lite models do have a free tier, but content on the free tier is marked as used to improve Google's products.
Are thinking tokens billed on the Gemini API?
Yes. Output prices include thinking tokens on Argon and the Flash models, and Google says every token the model processes is counted and billed, including thinking tokens and tool instructions when using grounding. On Argon after 1 December, that means $20.00 per million.
How much does grounding with Google Search cost?
For Gemini 3 and newer models, 5,000 search requests a month are free across all those models, then $14 per 1,000, with a cap of 1,500 requests a day. One prompt can trigger several searches, and each is billed. That price is not changing on 1 December.
Does the Batch API make Gemini cheaper?
Google's paid-tier summary lists the Batch API at a 50% cost reduction. If that applies to Argon after 1 December, a workload costing $80.00 at Standard rates would cost $40.00 on Batch. Check the Batch tab on the pricing page for the exact post-December rate before relying on it.
Does this change what I pay for the Gemini app?
No. API prices apply to developers billed per token. Gemini app users pay a flat Google AI subscription. Their separate change is that Google AI Plus loses Pro model selection on 13 November 2026, while keeping its $4.99 monthly price in the US.
