Cloud Machine Learning Engine

Build superior models and deploy them into production

Try It Free

Focus on models, not operations

Google Cloud Machine Learning (ML) Engine is a managed service that enables developers and data scientists to build and bring superior machine learning models to production. Cloud ML Engine offers training and prediction services, which can be used together or individually. Cloud ML Engine is a proven service used by enterprises to solve problems ranging from identifying clouds in satellite images, to ensuring food safety, and responding four-times faster to customer emails.


Machine learning involves training a computer model to find patterns in data. The more high-quality data that you train a well-designed model with, the more intelligent your solution will be. You can build state-of-the-art model architectures with the TensorFlow deep learning framework that powers many Google products, from Google Photos to Google Cloud Speech. Cloud ML Engine enables you to automatically design and evaluate model architectures to achieve an intelligent solution faster and without experts. Use TensorFlow Estimators for powerful distributed training, Keras to easily build custom estimators, or low-level TensorFlow for full control. Cloud ML Engine scales to leverage all your data. It can train any TensorFlow model at large scale on a managed cluster.


Prediction incorporates intelligence into your applications and workflows. Once you have a trained model, prediction applies what the computer learned to new examples. ML Engine offers two types of prediction:

Online Prediction deploys ML models with serverless, fully managed hosting that responds in real time with high availability. Our global prediction platform automatically scales to adjust to any throughput. It provides a secure web endpoint to integrate ML into your applications.

Batch Prediction offers cost-effective inference with unparalleled throughput for asynchronous applications. It scales to perform inference on TBs of production data.

Deploy multiple frameworks

Online Prediction enables developers and data scientists to seamlessly deploy ML models into production—no Docker container required. Users can import models that have been trained anywhere.

Cloud Machine Learning Engine Features

Automatic Resource Provisioning
Focus on model development and deployment without worrying about infrastructure. The managed service automates all resource provisioning and monitoring. Build models using managed distributed training infrastructure that supports CPUs, GPUs, and TPUs. Accelerate model development, by training across many nodes, or running multiple experiments in parallel.
Achieve superior results faster by automatically tuning deep learning hyperparameters with HyperTune. Data scientists can manage thousands of tuning experiments on the cloud. This saves many hours of tedious and error prone work.
Portable Models
Use the open source TensorFlow SDK to train models locally on sample data sets and use the Google Cloud Platform for training at scale. Models trained using Cloud Machine Learning Engine can be downloaded for local execution or mobile integration. Also, import scikit-learn, XGBoost, Keras, and TensorFlow models that have been trained anywhere for fully-managed, real time prediction hosting—no Docker container required.
Server-Side Preprocessing
Push deployment preprocessing to Google Cloud with scikit-learn pipelines and tf.transform. This means that you can send raw data to models in production, and reduce local computation. This also prevents data skew being introduced through different preprocessing in training and prediction.
Google services are designed to work together. It works with Cloud Dataflow for feature processing and Cloud Storage for data storage.
Multiple Frameworks
Online Prediction supports multiple frameworks to serve classification, regression, clustering, and dimensionality reduction models.
  • scikit-learn for the breadth and simplicity of classical machine learning
  • XGBoost for the ease and accuracy of extreme gradient boosting
  • Keras for easy and fast prototyping of deep learning
  • TensorFlow for the cutting edge power of deep learning

“ Google Cloud Machine Learning Engine enabled us to improve the accuracy and speed at which we correct visual anomalies in the images captured from our satellites. It solved a problem that has existed for decades. It will allow Airbus Defence and Space to continue to provide unrivaled access to the most comprehensive range of commercial Earth observation data available today ”

— Mathias Ortner Data Analysis & Image Processing Lead, Airbus Defense & Space

Cloud Machine Learning Engine Pricing

Cloud Machine Learning Engine charges for training ML models and running predictions with trained models. For detailed pricing information, please view the pricing guide.

Training - Predefined scale tiers - price per hour Training - Machine types - price per hour Batch prediction - price per node hour. Online prediction - price per node hour.
BASIC standard
STANDARD_1 large_model
PREMIUM_1 complex_model_s
BASIC_GPU complex_model_m
BASIC_TPU (Beta) complex_model_l
CUSTOM If you select CUSTOM as your scale tier, you have control over the number and type of virtual machines used for your training job. See the table of machine types. standard_gpu
standard_p100 (Beta)
complex_model_m_p100 (Beta)
cloud_tpu (Beta)
If you pay in a currency other than USD, the prices listed in your currency on Cloud Platform SKUs apply.