Skip Navigation

Machine Learning Services pricing

Bring AI to everyone with an end-to-end, scalable, trusted platform

The Azure Machine Learning Experimentation and Model Management services offer a cloud, on-premises and edge solution for data scientists and developers to bring AI everywhere, to everyone. To learn more about the suite of offerings available in Machine Learning Studio, visit the Machine Learning Studio pricing page.

Pricing details

The below pricing reflects a preview discount.

Experimentation pricing

Standard
Price First 2 seats—Free
Seats 3 and above—$- per seat per month

Model Management pricing

Dev/Test Standard S1 Standard S2 Standard S3 *
Tier Price per month $- $- $- $-
Features
Managed Models 20 100 1,000 10,000
Managed Deployments 2 10 100 1,000
Available Cores** 4 16 120 800
*If you require more than the available quantities of managed models, managed deployments and/or available cores that are included in the S3 tier then you can purchase multiple units of S3. For details, please refer to the FAQ section below. **Denotes the number of cores that can be used for deployments at any given time. Does not include charges for compute-hours. For details, please refer to the FAQ section below.

Support & SLA

  • We provide technical support for all Azure services released to general availability, including Machine Learning standard tier, through Azure Support starting at $29/month. Billing and subscription management support is provided at no cost.
  • Technical support for the Machine Learning free tier is only available through community forums. Training videos and documentation are also available to support the user community.
  • SLA—For the Request Response Service (RRS), we guarantee 99.95% availability of API transactions. For the Batch Execution Service (BES) and management APIs, we guarantee 99.9% availability of API transactions. We do not provide an SLA for Machine Learning free tier. To learn more about the SLA, please visit the SLA page.

FAQ

Azure Machine Learning Workbench

  • No. Azure Machine Learning Workbench is a free application. You can download it on as many machines and for as many users, as you need. To use the Azure Machine Learning Workbench, you must have an Experimentation account.

Azure Machine Learning Experimentation Service

  • Each seat is an Azure user that is added to the Experimentation account. The first two seats in your subscription are free. However, free seats and Dev/Test pricing will not apply to free or trial subscriptions, nor to subscriptions deriving from any other Azure offer.

  • No, the Experimentation Service allows as many experiments as you need and charges only based on the number of users. Experimentation compute resources are charged separately.

  • The Azure Machine Learning Experimentation service can execute your experiments on the following—local machine (direct or Docker-based), Azure compute resources (virtual machines) and HDInsight. It also needs to access an Azure Blob Storage account to store tracked execution outputs. It can also optionally use a Visual Studio Team Service account for version-controlling your project using a Git repo. Please note that you will be billed independently for any consumed compute and storage resources, based upon their individual pricing.

Azure Machine Learning Model Management

  • Azure Machine Learning models can currently be used with Azure IoT Edge at no charge.

  • No. Web services can be called as often as required, without any Model Management billing implications. You have full control to scale your deployments to meet the needs of your applications.

  • A model is the output of a training process and is the application of a machine learning algorithm to training data. The Model Management service enables you to deploy models as web services, manage various versions of the models and monitor the performance of your models and associated metrics. Managed models are models that have been registered with an Azure Machine Learning Model Management account. As an example, consider a scenario where you are trying to forecast sales. During the experimentation phase, you generate many models by using different datasets or algorithms. In a case where you have generated four models with varying accuracies, you may choose to register only the model with the highest accuracy.

    Any time you register a new model or register a new version of an existing model, it gets counted as part of your plan. At any point in time, you may have up to the maximum number of managed models denoted by the tiers you have purchased.

  • The Model Management service allows you to deploy models as packaged web service containers in Azure that can be invoked using REST APIs. Each web service is counted as a single deployment and the total number of active deployments running are counted towards your plan. At any point in time you may have up to the maximum number of deployments denoted by the tier you have purchased. Using the sales forecasting example, by deploying your best performing model, you will increment your plan with one deployment. If you then retrain and redeploy your model you will have two deployments. If you determine that the newer model is better and delete the original, your deployment count will decrement by one.

  • The Azure Machine Learning Model Management can run your deployments as docker containers on the Azure Container Service, Azure Virtual Machines and local machines with more targets coming in the future. Please note that you will be billed independently for any consumed compute resources, based upon their individual pricing.

  • The Azure Machine Learning Model Management service provides enhanced capabilities to optimise deployment on large clusters. You can deploy and manage models up to the total number of cores deployed across compute resources that you have provisioned. For example, if you have deployed an Azure Container Service cluster using 2 master nodes of D13 VMs (8 cores per node)and 10 worker nodes of D13 VMs (8 cores per node), your total number of cores is (2 x 8) + (10 x 8) = 96.

  • Only one DEV/TEST unit can be allocated per Azure subscription, but multiple units of S1, S2 and S3 can be combined. For instance, if you would like to have 25 managed deployments, you can purchase 3 units of Model Management S1.

  • You can change the number of units, either up or down, using either the Azure Management Portal or the CLI.

  • You get the best experience when you deploy models created using the Experimentation Service, but the models that you can deploy are not limited to the ones created using Experimentation Service. We support a variety of models (such as Spark ML, TensorFlow, CNTK, scikit-learn, Keras, etc.) created using tools such as Azure Batch AI Training, Microsoft ML Server or any other third-party tools.

  • You will be billed daily. For billing purposes, a day commences at midnight UTC. Bills are generated monthly. As a specific example, let’s say you subscribe to the Experimentation service for a team of 10 users. You have also purchased 3 units of the S1 Model Management tier.

    • Experimentation account charges—(((seats * days) – included) * dailyrate)
    • 2 free seats * 31 days = 62 seatdays included free each month, per subscription
    • Model Management account charges—(units * days * tierdailyrate)

    For a billing month of 30 days:

    • Experimentation account charges: (((10 * 30) – 62) * dailyrate)
    • Model Management account charges: (3 * 30 * tierdailyrate)

    Please note that you will incur separate charges for any Azure services consumed in conjunction with Azure Machine Learning, including but not limited to compute charges, HDInsight, Azure Container Service, Azure Container Registry, Azure Blob Storage, Application Insights, Azure Key Vault, Visual Studio Team Services, Virtual Network, Azure Event Hub and Azure Stream Analytics.

  • No. Azure Machine Learning Packages for Computer Vision, Text Analytics and Forecasting are free Python Packages. You can download them on as many machines and for as many users, as you need. Learn more about AML Packages.
For more information about pricing, please refer to the documentation FAQ.

Resources

Estimate your monthly costs for Azure services

Review Azure pricing frequently asked questions

Learn more about Machine Learning Services

Review technical tutorials, videos, and more resources

Added to estimate. Press 'v' to view on calculator View on calculator

Learn and build with $200 in credit and keep going for free