Prices for usage of AutoML Tables are computed based on the underlying GCP resources required for model training, model deployment, batch prediction, and online prediction.
Prices are listed in US Dollars (USD). If you pay in a currency other than USD, the prices listed in your currency on Cloud Platform SKUs apply.
These prices reflect a Beta program discount and are subject to change at General Availability.
Free trial: You can try AutoML Tables for free by using 6 free node hours each for training and for batch prediction, per billing account. Your free node hours are issued right before you create your first model, and you have up to one year to use them.
Model training costs
Model training costs $19.32 per hour of compute resources used to train your model, billed at the granularity of seconds. This price includes the use of 92 n1-standard-4 equivalent machines in parallel.
If training fails for any reason other than a user-initiated cancellation, you are not billed for training time. If you cancel the training operation, you are charged for any training time used before the operation is cancelled. If the cancellation does not succeed, you are charged for the entire training time.
Model deployment costs
Model deployment costs $0.005 per GiB per hour per machine that a model is deployed, billed at the granularity of MiB per second. We currently replicate your model to memory in 9 machines for low latency serving purposes, so there is a 9x multiplier applied to this cost.
For example, if the size of your model is 10 GiB, and you deploy for 3 hours, you would be charged $0.005 * 10 * 3 * 9, or $1.35.
Models must be deployed before they can be used to provide online predictions.
Batch prediction costs
Batch prediction using your model costs $1.16 per hour of compute resources used, billed at the granularity of seconds. This price includes the use of 5.5 n1-standard-4 equivalent machines in parallel.
Online prediction costs
Online predictions using your model cost $0.21 per hour of compute resources used, aggregated across all requests, and billed at the granularity of milliseconds. This price covers the 1 n1-standard-4 equivalent machine used to process the prediction. It does not include the cost of deploying the model.