Skip to main content

​
Parsing Models Overview

A Document Pre-Trained Transformer (DPT) is the model that powers the parsing capabilities of the ADE Parsing APIs. The DPT identifies document layouts and chunks, then generates descriptive explanations (captions) for those chunks.

​
Availability

The ability to select a Document Pre-Trained Transformer is available:

​
Model Versions and Snapshots

The following table lists the available model values for the ADE Parse and ADE Async Parse API:
The dpt-1 and dpt-2-mini models are deprecated. Use DPT-3 on ADE Gen2 instead (see the Migration Guide), or use dpt-2 to stay on ADE Gen1.
Organizations on subscription and enterprise plans started moving to dpt-2-20260903 in batches on October 7, 2026, and every plan tier moved on October 8, 2026. Every DPT-2 model value, including dpt-2, dpt-2-latest, and every pinned snapshot, now runs on dpt-2-20260903, and responses report dpt-2-20260903 as the model version. For details, go to dpt-2-20260903.

​
Why Model Versioning Matters

When integrating the API, you have two options for specifying the model:
  1. Use a general model name (like dpt-2 or dpt-2-latest) to always get the newest version. This automatically give you improvements and updates, but parsing results may change when new model versions are released
  2. Use a specific snapshot (like dpt-2-20260903) to pin to an exact model version. This keeps parsing results consistent until the snapshot is retired or migrated, but you won’t receive improvements.
If you use only a general model name like dpt-2 in production, your application may produce different results when we release model updates. Consider whether you need consistent results or prefer to receive the latest improvements.

​
Understanding Snapshots and -latest

Snapshots are versions of a model released on specific dates. A snapshot keeps the same parsing behavior until it is retired or migrated to a newer snapshot. For the current migration, go to dpt-2-20260903. The latest suffix points to the current default snapshot of that model. The default is not always the newest snapshot. For example, dpt-2-latest pointed to an earlier snapshot until your organization moved to dpt-2-20260903.

​
DPT-2

DPT-2 was introduced in September 2025. It builds upon an earlier model, and offers these advanced features:
  • Agentic Table Captioning: DPT-2 can parse large, complex, no-gridline, and merged-cell tables with unprecedented fidelity. Every cell is preserved, aligned, and made accessible—enabling cell-level grounding so you know exactly where values came from.
  • Refined Figure Captioning: Logos, seals, and small figures are now identified precisely and concisely, eliminating the noise of verbose descriptions.
  • Smarter Layout Detection: Fewer chunks are missed, even in messy scans. DPT-2 can even detect stamps inside tables and process them separately—critical for compliance workflows.
  • Expanded Chunk Ontology: Beyond text, tables, and figures, DPT-2 now recognizes attestation (signatures, stamps, seals), ID cards, logos, barcodes, and QR codes—ensuring all document elements are classified consistently. To learn more, go to Chunk Types.

​
dpt-2-20260903

The dpt-2-20260903 snapshot runs DPT-2 on an updated underlying model. One of our third-party model providers is retiring the model that earlier snapshots use, so every plan tier moved to this snapshot.
  • Output close to earlier snapshots: The snapshot is tuned to match earlier DPT-2 output. Some documents return different text or table formatting.
  • No confidence scores: The confidence field in the Parse response is null.
Your endpoints and request parameters do not change.

​
When Organizations Moved

Organizations on subscription and enterprise plans started moving to dpt-2-20260903 in batches on October 7, 2026. Every plan tier moved on October 8, 2026. Since then, dpt-2, dpt-2-latest, and every pinned DPT-2 snapshot run on dpt-2-20260903.

​
DPT-2 Availability

The DPT-2 can be used in these API endpoints:

​
Set the Model in the API

When calling the ADE Parse or ADE Parse Jobs endpoint, you can set the model using the model parameter. If you omit the model parameter, the API will use the latest snapshot of the dpt-2 model. For example, run the command below to use the latest snapshot of DPT-2.

​
Set the Model with the Library

When using the ade-python library, you can set the model using the model parameter in the parse() function. If you omit the model parameter, the library will use the latest snapshot of the dpt-2 model. For example, use this code to parse a document with the latest snapshot of DPT-2:

​
Set the Model in the Playground

The Playground’s Model drop-down defaults to DPT-3 Pro, the ADE Gen2 model. To select a DPT-2 parsing model instead:
  1. Open a project in the Playground.
  2. Click the Parse tab.
  3. In the top-right corner of the parse result panel, click Config.
  4. From the Model drop-down, select the model you want to use.
  5. Click Re-run all files. Configure Parse Settings
Switching the model re-parses every file in the project. If you switch back to a model that was already applied to the project, the Playground reuses the cached parse results instead of re-parsing.