Google Vertex AI Pricing Calculator

1. Pick a model and a service tier. The three price fields fill with the figures Google listed on 2026-10-02. Edit any price that has changed since then.
2. Enter the input tokens and output tokens for one typical request, then your requests per day and active days per month.
3. Add the optional parts: the cached share of your input, your grounding queries, and a monthly budget.

What each result tells you:

The token counts are your own numbers. Google reports the real counts in the usage metadata of every response, which is the best place to read them.

Google Vertex AI pricing for Gemini models is pay-as-you-go billing per 1 million tokens, with separate prices for input, cached input and output. A token is a small piece of text. Google's pricing page says about 4 characters make 1 text token, which is an approximation and not an exact rule.

The product has a new name. Google now calls it Gemini Enterprise Agent Platform (formerly Vertex AI), and the official pricing page is titled Agent Platform Pricing. This page keeps the older name because many people still search for it. The prices below are the ones on that official page.

Three things set the bill: the model, the number of tokens, and the service tier. Google lists Standard, Priority and Flex or Batch tiers for Gemini models. Extra features such as Grounding with Google Search are billed on top.

The prices Google listed on 2026-10-02 for the global endpoint, in US dollars per 1 million tokens:

This calculator covers Gemini text generation only. It does not price image, audio or video output, model tuning, or other Google Cloud services. Image output has its own price list and is covered by a separate tool in the related tools below.

The calculator turns one request into a month of tokens, then prices each token type at its own rate.

1. Requests per month = requests per day x active days.
2. Input tokens per month = input tokens per request x requests per month.
3. Cached tokens = input tokens per month x cached share. Fresh input tokens are the rest.
4. Output tokens per month = output tokens per request x requests per month.
5. Each token total is divided by 1,000,000 and multiplied by its price.
6. Grounding cost = (queries - free allowance) / 1,000 x the grounding price. It is never below zero.

Google charges only for requests that return a 200 response code. Failed requests with 4xx or 5xx codes are not charged for input or output, so count successful calls. A quick way to turn text length into tokens is the AI token calculator.

A support chat app runs on Gemini 3.5 Flash at the Standard tier. Each request has 1,500 input tokens and 500 output tokens. The app serves 2,000 requests a day for 30 days.

Meaning: the app costs about $405.00 a month, or $4,860.00 a year at steady traffic. Output is one quarter of the tokens but 66.7% of the bill, because output costs six times as much as input on this model.

Model Choice

The model sets the price per token. The example workload costs $405.00 on Gemini 3.5 Flash and $102.00 on Gemini 3.5 Flash-Lite. A cheaper model only saves money when its answers are good enough for the task.

Output Length

Output tokens cost more than input tokens on every model in the table. Google prices the response and the reasoning tokens together as output, so a model that reasons at length raises this line.

Long Context

Gemini 3.1 Pro Preview and Gemini 2.5 Pro have a second price above 200K input tokens. Google states that all tokens, input and output, are then charged at long context rates. A thousand requests of 250,000 input tokens cost $1,000.00 in input on Gemini 3.1 Pro Preview, not $500.00.

Context Caching

Google's caching guide gives a 90% discount on cached tokens for Gemini 2.5 and later models. With 60% of the example input served from cache, the bill falls from $405.00 to $332.10. Explicit caches also carry a storage charge that this tool does not add.

Service Tier

Flex or Batch prices are half the Standard prices in the table. The example falls to $202.50 a month. Google's batch guide says the cache and batch discounts do not stack.

Endpoint Region

Google lists a higher price for non-global endpoints on Gemini 3 and later models from July 1, 2026. Gemini 3.5 Flash is $1.65 input and $9.90 output there, which lifts the example to $445.50.

Grounding

Grounding with Google Search is billed per query above the free allowance. With 60,000 queries a month on a Gemini 3 model, 55,000 are billable and add $770.00.

The cheapest model in the table costs $21.00 a month for the example workload, and the most expensive costs $540.00. Every row uses 90 million input tokens and 30 million output tokens at Standard global prices.

Gemini 3.8 Flash shows its introductory price. Google states that standard pricing of $1.50 input and $7.50 output per 1 million tokens applies from January 1, 2027. The pricing page also notes that this promotional price is delivered as credits back on net spend.

What Should You Budget?

Budget for the tokens you measure, plus headroom for growth. Run a small test, read the token counts from the response metadata, and enter the averages here. A team that pays for several AI services can total them with the AI tool stack cost calculator.

Before Launch

A product team enters the expected traffic and token sizes to see whether the feature fits its budget before any code ships.

Choosing Between Models

A developer keeps the workload fixed and switches the model. The difference in monthly cost shows what a stronger model has to be worth.

Planning a Batch Pipeline

A data team that can wait for results compares the Standard tier with Flex or Batch. Google says most batch jobs finish within 24 hours.

Checking Response Speed Against Cost

A chat product needs fast replies as well as a low bill. The token generation speed simulator shows how long a response of a given size takes to appear.

1. Counting Requests Instead of Tokens

Google bills tokens, not calls. One request with 10,000 input tokens costs the same input as 20 requests with 500 tokens each.

2. Leaving Out Output

Output is the larger share in many chat workloads. In the example it is 66.7% of the bill.

3. Forgetting the Conversation History

A chat app resends earlier turns with each request. Input tokens per request grow as the conversation gets longer, so use an average across the whole session.

4. Ignoring Reasoning Tokens

Reasoning tokens are billed as output. A short visible answer can still carry a large output count.

5. Using an Old Price List

Model names and prices change often. Older Gemini 1.5 figures no longer describe the current models, so check the date on any price you reuse.

6. Applying the Cache Price to All Input

Only tokens that hit the cache get the lower price. Set the cached share to the part of your prompt that repeats.

The arithmetic is exact for the prices and token counts you enter. The estimate is only as good as those inputs, and real traffic varies from day to day.

No. Google now names the product Gemini Enterprise Agent Platform, and its own product page adds the words formerly Vertex AI. The official pricing page is titled Agent Platform Pricing.

It depends on the model. On 2026-10-02 the Standard global prices ran from $0.10 input and $0.40 output for Gemini 2.5 Flash Lite to $2.00 input and $12.00 output for Gemini 3.1 Pro Preview.

No. Google states that you are charged only for requests that return a 200 response code. Requests that return 4xx or 5xx codes are not charged for input or output.

There is no fixed number. Google says about 4 characters make 1 text token, so a short English word is often 1 or 2 tokens. Read the real count from the usage metadata in each response.

Reasoning tokens are billed at the output price. Google lists the output line as text output for the response and reasoning, so they are part of the output token count.

For Gemini 3 models, Google includes 5,000 grounding queries a month and bills $14 per 1,000 queries above that. For Gemini 2.5 models it lists $35 per 1,000 grounded prompts above a daily free allowance.

In the table on this page, every Flex or Batch input and output price is half the Standard price. Google does not let the batch and cache discounts stack, and batch results can take up to 24 hours.

Cached input tokens cost 90% less than standard input tokens on Gemini 2.5 and later models. The saving applies only to the tokens that hit the cache, and explicit caches add a storage charge.

Yes, for newer models. From July 1, 2026, Google lists higher prices for non-global endpoints on Gemini 3 and later models. Gemini 3.5 Flash is $1.65 input there instead of $1.50.

Real requests vary in size, some hit the cache and some do not, and traffic is uneven. The invoice also includes other Google Cloud services, taxes and any credits.

No. The calculation runs in your browser and nothing is sent to our servers. Anything you Save stays in this browser only.

• Agent Platform pricing: Generative AI (Google Cloud, official price list, checked 2026-10-02).
• Gemini Enterprise Agent Platform (formerly Vertex AI) (Google Cloud, product page showing the new name).
• Context caching overview (Google Cloud documentation).
• Batch inference with Gemini (Google Cloud documentation).