# Known limitations
> This page provides an overview of the current specifications and parameters for optimal utilization of Document UnderstandingTM modern projects.
This page provides an overview of the current specifications and parameters for optimal utilization of Document UnderstandingTM modern projects.
## Project-based limits
Description
Limit
Supported file formats
PNG
JPG/JPEG
PDF
TIF/TIFF
Image size limits
Minimum: 50 x 50 pixels
Maximum: 10,000 x 10,000 pixels
Maximum number of pages in a document type
5000 After a certain point, adding more data doesn't improve model performance. This limitation makes sure that you do not annotate more documents than needed.
Maximum number of pages in a document
500 for previous generation models, 100 for Helix Extractor models
Maximum number of pages in a project
30,000
Maximum number of pages per document for pre-labeling
20 If a documents has more than 20 pages, only the first 20 pages are pre-annotated.
Maximum number of fields
300
Maximum number of requests processed in parallel
10
Maximum number of pages for out-of-the-box model processing
20
Maximum file size for digitization
160 MB
Maximum pages per document for digitization
500 pages
Number of characters for the classification name
Minimum: 1
Maximum: 50
Number of characters for the classification description
Minimum: 0
Maximum: 2000
Number of characters in a document for classification
Minimum: 0
Maximum: 1e10 (virtually unlimited)
Number of of classifiers
Minimum: 1
Maximum: 50
Maximum number of pages processed by UiPath Helix Extractor models at run-time
100 pages
## Activity or API-based limits
Description
Limit
Supported file formats
PNG
JPG/JPEG
PDF
TIF/TIFF
Image size limits
Minimum: 50 x 50 pixels
Maximum: 10,000 x 10,000 pixels
Maximum file size for digitization
160 MB
Maximum number of requests processed in parallel (a project version deployed in a modern project)
10
Generative extraction in activities
Minimum number of characters in a document: 10
Document length: 500 pages
Fields: 150
Prompt length: 1000 characters
Generative classification in activities
Document length: 500 pages
Prompt length: 1000 characters
Maximum number of pages for out-of-the-box model processing
20 pages
## Current limitations when migrating existing classic projects
* Currently, importing datasets larger than 5000pages is not supported. Only the initial 5000 pages will be successfully imported, with any additional pages failing to do so. For example, if your dataset consists of 4999 pages and you try to import a document of 4 pages, the process will not succeed.
* Batch names and corresponding batch results are not currently available. If your data is organized into batches, this information is not displayed yet, but it is saved.
* Exports from AI Center are not supported. Only exports from Document Manager are supported.
## Other limitations
* The Project Performance dashboard displays documents processed by consuming the respective project, either through APIs or through activities from the **DocumentUnderstanding.Activities** package.
* The Project Performance dashboard is not accessible in the United Kingdom and India due to Insights dashboards unavailability.
* When training a custom model on version 24.4 or 24.10 in a modern project, data extraction may fail if the model is trained on a single document template. This occurs because the model can become overfit to one layout, causing extraction errors even on visually similar documents. You can improve accuracy by training the model with a more diverse dataset.
* When using the Project Extractor activity, the **Project Name** must be provided with the exact same letter casing as it was originally defined in the Document Understanding project.
* The Document Type Classifier does not support documents written in non-Latin alphabets, such as Hebrew, Chinese or Japanese. When document type classification is used with such documents, pipelines may fail or produce unexpected results, including encoding-related errors. Document extraction can still work with non-Latin languages if classification is not used.