Custom text classification limits

Use this article to learn about the data and service limits when using custom text classification.

Language resource limits

  • Your Language resource has to be created in one of the supported regions and pricing tiers listed below.

  • You can only connect 1 storage account per resource. This process is irreversible. If you connect a storage account to your resource, you cannot unlink it later. Learn more about connecting a storage account

  • You can have up to 500 projects per resource.

  • Project names have to be unique within the same resource across all custom features.

Pricing tiers

Custom text classification is available with the following pricing tiers:

Tier Description Limit
F0 Free tier You are only allowed one F0 tier Language resource per subscription.
S Paid tier You can have unlimited Language S tier resources per subscription.

See pricing for more information.

Regional availability

Custom text classification is only available in some Azure regions. Some regions are available for both authoring and prediction, while other regions are prediction only. Language resources in authoring regions allow you to create, edit, train, and deploy your projects. Language resources in prediction regions allow you to get predictions from a deployment.

Region Authoring Prediction
Australia East
Brazil South
Canada Central
Central India
Central US
East Asia
East US
East US 2
France Central
Japan East
Japan West
Jio India West
Korea Central
North Central US
North Europe
Norway East
South Africa North
South Central US
Southeast Asia
Sweden Central
Switzerland North
UAE North
UK South
West Central US
West Europe

API limits

Item Request type Maximum limit
Authoring API POST 10 per minute
Authoring API GET 100 per minute
Prediction API GET/POST 1,000 per minute
Document size -- 125,000 characters. You can send up to 25 documents as long as they collectively do not exceed 125,000 characters


If you need to send larger files than the limit allows, you can break the text into smaller chunks of text before sending them to the API. You use can the chunk command from CLUtils for this process.

Quota limits

Pricing tier Item Limit
F Training time 1 hour per month
S Training time Unlimited, Pay as you go
F Prediction Calls 5,000 text records per month
S Prediction Calls Unlimited, Pay as you go

Document limits

  • You can only use .txt. files. If your data is in another format, you can use the CLUtils parse command to open your document and extract the text.

  • All files uploaded in your container must contain data. Empty files are not allowed for training.

  • All files should be available at the root of your container.

Data limits

The following limits are observed for the custom text classification.

Item Lower Limit Upper Limit
Documents count 10 100,000
Document length in characters 1 128,000 characters; approximately 28,000 words or 56 pages.
Count of classes 1 200
Count of trained models per project 0 10
Count of deployments per project 0 10

Naming limits

Item Limits
Project name You can only use letters (a-z, A-Z), and numbers (0-9) ,symbols _ . -,with no spaces. Maximum allowed length is 50 characters.
Model name You can only use letters (a-z, A-Z), numbers (0-9) and symbols _ . -. Maximum allowed length is 50 characters.
Deployment name You can only use letters (a-z, A-Z), numbers (0-9) and symbols _ . -. Maximum allowed length is 50 characters.
Class name You can only use letters (a-z, A-Z), numbers (0-9) and all symbols except ":", $ & % * ( ) + ~ # / ?. Maximum allowed length is 50 characters.
Document name You can only use letters (a-z, A-Z), and numbers (0-9) with no spaces.

Next steps