Edit

Share via

Azure Content Understanding in Foundry Tools service quotas and limits

This article offers a reference of the quotas and limits for the Azure Content Understanding in Foundry Tools service.

General limits

Property Limit
Analyzer ID 1-64 characters. Alphanumeric, period, and underscore. Pattern: [a-zA-Z0-9._]{1,64}
URL properties ≤ 8,192 characters
Description properties ≤ 1,024 characters
Field names ≤ 64 characters. Unicode letters, numbers, combining marks, connecting punctuation, period, hyphen, and underscore. Pattern: [\p{L}\p{Nl}\p{Mn}\p{Mc}\p{Nd}\p{Pc}._-]{1,64}
Tags ≤ 10 tags
Tag key ≤ 64 characters. Alphanumeric and + - . : = _ / characters. Pattern: [a-zA-Z0-9+-.:=_/]{1,64}
Tag value ≤ 256 characters. Alphanumeric and + - . : = _ / characters. Can be empty. Pattern: [a-zA-Z0-9+-.:=_/]{0,256}
Image reference ID ≤ 256 characters

Resource limits

Quota Standard (S0)
Max analyzers 100k
Max analysis/min 1000 pages/images
Four hours of audio
Four hours of video
Max operations/min 3000

Supported generative models

Content Understanding connects to Foundry Models for generative capabilities. The service is periodically updated to add support for more models. To learn more see Connect your Content Understanding analyzer to Foundry model deployments

The currently supported models are:

Model Type Model Version
Chat Completion gpt-4o 2024-08-06
Chat Completion gpt-4o 2024-11-20
Chat Completion gpt-4o-mini 2024-11-20
Chat Completion gpt-4.1 2024-11-20
Chat Completion gpt-4.1-mini 2024-11-20
Chat Completion gpt-4.1-nano 2024-11-20
Embeddings text-embedding-3-small
Embeddings text-embedding-3-large
Embeddings text-embedding-ada-002

Analyzers

Input file limits

Document and text

Supported File Types File Size Length
✓ .pdf
✓ .tiff
✓ .jpg, .jpeg, .jpe, .png, .bmp, .heif, .heic		 ≤ 200 MB ≤ 300 pages
✓ .docx, .xlsx, .pptx ≤ 200 MB ≤ 1M characters
✓ .txt
✓ .html, .md, .rtf
✓ .eml, .msg
✓ .xml		 ≤ 1 MB ≤ 1M characters

Note

Pro mode (preview) currently only supports .pdf, .tiff, and image file types as input. Total input may not exceed 100 MB and 150 pages.

Image

Supported File Types File Size Resolution
✓ .jpg, .jpeg, .jpe, .png, .bmp, .heif, .heic ≤ 200 MB Min: 50 x 50 pixels
Max: 10k x 10k pixels

Audio

Supported File Types File Size Length
✓ .wav (PCM, A-law, μ-law)
✓ .mp3
✓ .mp4
✓ .opus, .ogg (Opus)
✓ .flac
✓ .wma
✓ .aac
✓ .amr (AMR-NB, AMR-WB)
✓ .3gp (AMR-NB, AMR-WB)
✓ .webm (Opus, Vorbis)
✓ .m4a (AAC, ALAC)
✓ .spx		 Max: 300 MB† Max: 2 hours†

† Content Understanding supports audio files up to 1 GB and 4 hours in duration, but transcription time is substantially reduced for files ≤ 300 MB or ≤ 2 hours.

Video

Supported file types and resolution
Supported File Types Resolution
✓ .mp4, .m4v
✓ .flv (H.264 and AAC)
✓ .wmv, .asf
✓ .avi
✓ .mkv
✓ .mov		 Min: 320 x 240 pixels
Max: 1920 x 1,080 pixels
File size limits
Upload Method File Size Length Description
analyzeBinary API (direct upload) ≤ 200 MB ≤ 30 minutes Uploading video files directly in the API request body using the analyzeBinary API. This method is used by the Microsoft Foundry UX and Content Understanding Studio UX.
analyze API (file reference) Max: 4 GB Max: 2 hours Referencing video files via URL from Azure Blob Storage or similar storage when using the analyze API

Note

Video analysis has the following limitations:

  • analyzeBinary API: Maximum file size of 200 MB and maximum duration of 30 minutes when uploading video directly in the request body
  • Frame sampling: Analyzes approximately one frame per second, which may miss quick movements or brief events
  • Resolution: All frames are scaled to 512 x 512 pixels, which may affect visibility of small details or distant objects

Field schema limits

Content Understanding supports both basic field value types and nested structures, including lists, groups, tables, and fixed tables.

  • Basic field value types: string, date, time, number, integer, and boolean.
  • List field: A sequence of values of the same type, represented as an array of basic fields in the API.
  • Group field: A set of semantically related fields, represented as an object of basic fields in the API.
  • Table field: A variable number of items with fixed subfields, represented as an array of objects of basic fields in the API.
  • Fixed table field: A group of fields with shared subfields, represented as an object of objects of basic fields in the API.

Basic limits

Property Document Text Image Audio Video
Max fields 1000 1000 1000 1000 1000
Max classify field categories 300 300 300 300 300
Supported generation methods extract
generate
classify		 generate
classify		 generate
classify		 generate
classify		 generate
classify
  • The Max fields limit includes all named fields. For example, a list of strings counts as one field, while a group with string and number subfields counts as three fields.
  • The Max classify field categories limit is the total number of categories across all fields using the classify generation method.

Knowledge source limits

Type Limits
Training data Documents only
1 GB total
50k pages/images total

Segmentation/Classification

General limits

Note

This limit is for Content Understanding segmentation/classification itself, not classify fields within the extraction capability.

Property Limit
Category name Can't start with a dollar sign ($)
Category name and description Maximum 120 characters for combined name and description in each category
Number of categories 200 per analyzer for documents, 1 for videos
Hierarchical classification 5 layers for documents, 2 layers for videos
  • Last updated on