Skip to content
Calcrivo

Vertex AI Cost Calculator

Estimate Vertex AI spend across generative token usage, custom training node-hours and always-on prediction endpoints.

Inputs

thousands
tokens
tokens
USD/M tokens
USD/M tokens

Output tokens routinely cost several times input tokens, so response length matters more than prompt length.

%
%
hours
USD/node-hour
accelerators
USD/hour
nodes

An endpoint with a minimum replica count is billed whether or not it serves traffic.

hours
USD/node-hour

Monthly Cost

$974.70

Generative Token Charge

$447.50

Custom Training Charge

$206.00

Prediction Endpoint Charge

$321.20

Tokens Processed

575.00millions

Output Share of the Token Charge

69.8%

Cost per Thousand Requests

$1.9494

Largest Cost Line

Generative token usage — shorten responses before anything else

Step by step

  1. Values used

    Model requests per month = 500 thousands; Average input tokens per request = 900 tokens; Average output tokens per request = 250 tokens; Input price per million tokens = 0.3000 USD/M tokens; Output price per million tokens = 2.50 USD/M tokens; Input tokens served from context cache = 0 %; Discount on cached input tokens = 75 %; Custom training hours = 40 hours; Training price per node-hour = 0.1900 USD/node-hour; Accelerators attached during training = 2 accelerators; Accelerator price per hour = 2.48 USD/hour; Prediction endpoint nodes = 2 nodes; Hours the endpoint is deployed = 730 hours; Prediction price per node-hour = 0.2200 USD/node-hour

  2. Vertex AI Cost

    token cost = (fresh input tokens + cached input × (1 − cache discount)) ÷ 1M × input price + output tokens ÷ 1M × output price; training = hours × (node price + accelerators × accelerator price); prediction = nodes × hours × node price.

  3. Monthly Cost

    = 974.70

  4. Generative Token Charge

    = 447.50

  5. Custom Training Charge

    = 206.00

  6. Prediction Endpoint Charge

    = 321.20

  7. Tokens Processed

    = 575.00 millions

  8. Output Share of the Token Charge

    = 69.8

How it works

Vertex AI bills three unrelated things: tokens for generative models, node and accelerator hours for training, and deployed replica hours for prediction endpoints. Because output tokens cost several times input tokens, response length is usually the strongest lever on a generative bill, while a deployed endpoint with a minimum replica count is a fixed monthly floor regardless of traffic. Teams estimate inference cost from request counts and are then surprised by verbose outputs, or leave a training endpoint deployed after an experiment ends and pay for it all month. Confirm current per-token and per-node-hour rates for your model and region in the Google Cloud pricing calculator, since model prices change with every release.

Formula

Vertex AI Cost

token cost = (fresh input tokens + cached input × (1 − cache discount)) ÷ 1M × input price + output tokens ÷ 1M × output price; training = hours × (node price + accelerators × accelerator price); prediction = nodes × hours × node price.

output tokens
Generated tokens, priced several times higher than input tokens
context cache
Reused prompt prefix billed at a reduced rate
prediction node-hour
Charge for a deployed endpoint replica, billed while deployed rather than per request

Frequently Asked Questions

How is Vertex AI Cost calculated?

token cost = (fresh input tokens + cached input × (1 − cache discount)) ÷ 1M × input price + output tokens ÷ 1M × output price; training = hours × (node price + accelerators × accelerator price); prediction = nodes × hours × node price. Vertex AI bills three unrelated things: tokens for generative models, node and accelerator hours for training, and deployed replica hours for prediction endpoints. Because output tokens cost several times input tokens, response length is usually the strongest lever on a generative bill, while a deployed endpoint with a minimum replica count is a fixed monthly floor regardless of traffic.

Why does Vertex AI Cost matter?

Teams estimate inference cost from request counts and are then surprised by verbose outputs, or leave a training endpoint deployed after an experiment ends and pay for it all month. Confirm current per-token and per-node-hour rates for your model and region in the Google Cloud pricing calculator, since model prices change with every release.

What values do I need to enter?

This calculator takes 14 inputs: Model requests per month, Average input tokens per request, Average output tokens per request, Input price per million tokens, Output price per million tokens, Input tokens served from context cache, Discount on cached input tokens, Custom training hours, Training price per node-hour, Accelerators attached during training, Accelerator price per hour, Prediction endpoint nodes, Hours the endpoint is deployed, Prediction price per node-hour. The pre-filled defaults are a realistic starting point — replace them with figures from your own environment for a result you can act on.

Is an always-on endpoint cheaper than per-token billing?

Only at sustained high volume. A deployed endpoint bills replica hours whether or not requests arrive, so it wins when utilisation is high and steady, while token-billed managed models suit spiky or low-volume traffic because idle time is free.

You might also need