---
schema_version: "1"
id: 8e2b013d9d8f68c06faac69d25a7595c51f6d9406371085ed2f5fd8b9db061b1
title: Supported Models
collection: mulesoft
version: current
locale: en-us
portal_url: https://docs.salesforce.com/mulesoft/current/en-us/idp/supported-models.md
canonical_url: https://docs.mulesoft.com/idp/supported-models
source_site: docs.mulesoft.com
content_type: documentation
content_hash: f9e7d2d045119f5479b20878fc73ef5c91d03bc86927995a40d2184be819209a
description: MuleSoft Documentation Site
latest: true
---
---
product: IDP
---

# Supported Models

> For the full documentation index, see: https://docs.mulesoft.com/llms.txt

Einstein supports these predictive multimodal models:

- OpenAI’s `GPT-4o (gpt-4o)` LLM

- OpenAI’s `GPT-4o Mini (gpt-4o-mini)` LLM

- OpenAI’s `GPT-5.1` LLM

- OpenAI’s `GPT-5.2` LLM

- Google’s `GEMINI-3.5 Flash` LLM

- Anthropic’s `Claude Sonnet 4.6` LLM

- Anthropic’s `Claude Opus 4.6` LLM

> [!WARNING] **Einstein OpenAI GPT-4o and Einstein OpenAI GPT-4o Mini retire January 31, 2027**.
>
> To prevent service interruption, migrate actions to a supported model by January 31, 2027.

When creating a document action, you can select the model and extraction settings such as PII
masking or image recognition. Each model performs differently under various conditions, so select
the one that aligns with your specific requirements.

For additional details on GPT models' lifecycle and deprecation, see
[Microsoft Foundry Models lifecycle and support policy](https://learn.microsoft.com/en-us/azure/foundry/openai/concepts/model-retirements?view=foundry-classic&tabs=text).

This table compares the supported models by stability, image recognition performance, location
callout support, and how each model provides confidence scores. The **Notes** column includes
recommendations and limits for each model, such as supported languages and the maximum number of
fields.

<table id="supported-models-table"><colgroup><col> <col> <col> <col> <col> <col> <col></colgroup><thead><tr><th>Model Name</th><th>Description</th><th>Stability</th><th>Image Recognition</th><th>Location Callouts</th><th>Confidence Score</th><th>Notes</th></tr></thead><tbody><tr><td><div><div><p>Einstein OpenAI GPT-4o</p></div></div></td><td><div><div><p>Suitable for most tasks, performing well on documents in non-Latin languages. Can compare font sizes and identify certain font styles.</p></div></div></td><td><div><div><p>High</p></div></div></td><td><div><div><ul><li><p>Standard performance</p></li><li><p>Low accuracy on reading checkboxes in forms</p></li></ul></div></div></td><td><div><div><p>Not supported</p></div></div></td><td><div><div><p>Calculated</p></div></div></td><td><div><div><ul><li><p>Ask the model to not make up information to prevent hallucinations</p></li><li><p>List of <a href="https://help.openai.com/en/articles/8357869-how-to-change-your-language-setting-in-chatgpt#h_513834920e">languages supported</a> by OpenAI models</p></li></ul></div></div></td></tr><tr><td><div><div><p>Einstein OpenAI GPT-4o Mini</p></div></div></td><td><div><div><p>Fast and useful for focused tasks. Tends to exhibit lazy reasoning.</p></div></div></td><td><div><div><p>High</p></div></div></td><td><div><div><ul><li><p>Standard performance</p></li><li><p>Low accuracy on reading checkboxes in forms</p></li></ul></div></div></td><td><div><div><p>Not supported</p></div></div></td><td><div><div><p>Calculated</p></div></div></td><td><div><div><ul><li><p>Ask the model to not make up information to prevent hallucinations</p></li><li><p>List of <a href="https://help.openai.com/en/articles/8357869-how-to-change-your-language-setting-in-chatgpt#h_513834920e">languages supported</a> by OpenAI models</p></li></ul></div></div></td></tr><tr><td><div><div><p>Einstein OpenAI GPT-5.1</p></div></div></td><td><div><div><p>Suitable for most tasks. Improved instruction following and accuracy compared to GPT-4o.</p></div></div></td><td><div><div><p>High</p></div></div></td><td><div><div><ul><li><p>Medium accuracy on reading checkboxes in forms</p></li></ul></div></div></td><td><div><div><p>Not supported</p></div></div></td><td><div><div><p>Calculated</p></div></div></td><td><div><div><ul><li><p>Meets EU data residency requirements</p></li><li><p>Ask the model to not make up information to prevent hallucinations</p></li><li><p>List of <a href="https://help.openai.com/en/articles/8357869-how-to-change-your-language-setting-in-chatgpt#h_513834920e">languages supported</a> by OpenAI models</p></li></ul></div></div></td></tr><tr><td><div><div><p>Einstein OpenAI GPT-5.2</p></div></div></td><td><div><div><p>Suitable for documents with charts, tables, and complex layouts due to improved visual understanding.</p></div></div></td><td><div><div><p>High</p></div></div></td><td><div><div><ul><li><p>Medium accuracy on reading checkboxes in forms</p></li><li><p>Performs slightly better than earlier GPT models when reading checkboxes and tables</p></li></ul></div></div></td><td><div><div><p>Not supported</p></div></div></td><td><div><div><p>Calculated</p></div></div></td><td><div><div><ul><li><p>Does not currently meet EU data residency requirements<br>For updates on data residency requirements, see <a href="https://help.salesforce.com/s/articleView?id=ai.generative_ai_large_language_model_support.htm&amp;type=5">Salesforce’s Large Language Models Support</a></p></li><li><p>Ask the model to not make up information to prevent hallucinations</p></li><li><p>List of <a href="https://help.openai.com/en/articles/8357869-how-to-change-your-language-setting-in-chatgpt#h_513834920e">languages supported</a> by OpenAI models</p></li></ul></div></div></td></tr><tr><td><div><div><p>Einstein GEMINI-3.5 Flash</p></div></div></td><td><div><div><p>Fast and the strongest all-round performer in current testing. Handles structured documents and invoices, including non-English formats (localized dates, currency, and thousand separators). Leads in checkboxes and handwriting.</p></div></div></td><td><div><div><p>Standard</p></div></div></td><td><div><div><ul><li><p>Best of the tested models at reading checkboxes and handwriting in multimodal mode</p></li></ul></div></div></td><td><div><div><p>Supported</p></div></div></td><td><div><div><p>Self-reported</p></div></div></td><td><div><div><ul><li><p>Ask the model to not make up information to prevent hallucinations</p></li><li><p>List of <a href="https://support.google.com/gemini/answer/13575153?hl=en">languages supported</a> by Gemini models</p></li></ul></div></div></td></tr><tr><td><div><div><p>Einstein Anthropic Claude Opus 4.6</p></div></div></td><td><div><div><p>High-accuracy extraction on structured and semi-structured documents with complex or varied layouts, including invoices. Favors conservative, no-guess answers over made-up values.</p></div></div></td><td><div><div><p>Standard</p></div></div></td><td><div><div><ul><li><p>Performs well on handwritten documents</p></li><li><p>Medium accuracy on reading checkboxes in forms</p></li></ul></div></div></td><td><div><div><p>Not supported</p></div></div></td><td><div><div><p>Self-reported</p></div></div></td><td><div><div><ul><li><p>Supports up to 15 fields per document action</p></li><li><p>Ask the model to not make up information to prevent hallucinations</p></li><li><p>List of <a href="https://platform.claude.com/docs/en/build-with-claude/multilingual-support">languages supported</a> by Anthropic models</p></li></ul></div></div></td></tr><tr><td><div><div><p>Einstein Anthropic Claude Sonnet 4.6</p></div></div></td><td><div><div><p>Good for general document extraction tasks requiring a balance of accuracy and speed. Performs well on structured and semi-structured documents with complex or varied layouts.</p></div></div></td><td><div><div><p>High</p></div></div></td><td><div><div><ul><li><p>Performs better than Einstein OpenAI GPT-5.2 on handwritten documents</p></li><li><p>Medium accuracy on reading checkboxes in forms</p></li></ul></div></div></td><td><div><div><p>Not supported</p></div></div></td><td><div><div><p>Self-reported</p></div></div></td><td><div><div><ul><li><p>Supports up to 15 fields per document action</p></li><li><p>Ask the model to not make up information to prevent hallucinations</p></li><li><p>List of <a href="https://platform.claude.com/docs/en/build-with-claude/multilingual-support">languages supported</a> by Anthropic models</p></li></ul></div></div></td></tr></tbody></table>

## Retired Models

After a model retires, IDP temporarily routes requests for that model to a current model.
Eventually, IDP removes the retired model from the product and you can no longer select it. To avoid
errors or service interruptions, update your document actions to use a supported model as soon as
possible.

<table><colgroup><col> <col> <col></colgroup><thead><tr><th>Model Name</th><th>Retirement Date</th><th>Suggested Replacement</th></tr></thead><tbody><tr><td><div><div><p>Einstein GEMINI-2.0 Flash 001</p></div></div></td><td><div><div><p>June 1, 2026</p></div></div></td><td><div><div><p>Einstein GEMINI-3.5 Flash</p></div></div></td></tr><tr><td><div><div><p>Einstein GEMINI-2.5 Flash</p></div></div></td><td><div><div><p>October 16, 2026</p></div></div></td><td><div><div><p>Einstein GEMINI-3.5 Flash</p></div></div></td></tr></tbody></table>

## Confidence Scores by Model

Each model returns a confidence score for every extracted value. Depending on the model, the score
is calculated or self-reported. This applies to all document action types.

- **Calculated**: IDP derives the confidence score from model probabilities rather than
  self-reported estimates. This score reflects model certainty and is generally more stable, but it
  doesn’t guarantee that the extracted value is accurate.

- **Self-reported**: IDP prompts the model to rate its own confidence when model probabilities
  aren’t available. This score reflects the model’s self-assessment and is generally less stable
  than a calculated score, but it provides a baseline estimate.

Self-reported confidence scores become less reliable as document processing complexity increases.
Reliability depends not only on schema complexity, such as processing numerous fields or complex
tables, but also on document quality, prompt logic, and structural variability across document
groups. Low-quality scans, prompts that require data transformation alongside extraction, and
inconsistent field availability across uploads can all degrade model self-assessment. Fields that
are correctly returned as not found can show a low or 0% confidence score, which sends the document
to human review even though the extraction is correct.

IDP continues to refine how it requests self-reported confidence scores, which reduces how often
this happens. However, because the model estimates its own confidence, no level of complexity
guarantees consistent scores. Prompt instructions about confidence, such as
`Fields that return "Not found" must have a 99% confidence rate`, don’t reliably change the reported
score.

To get more reliable results with self-reported confidence scores:

- Keep the document action as simple as possible. Remove fields, tables, and prompts that aren’t
  required, and consider splitting large extractions into multiple document actions.

- Test the document action against a large, representative set of documents, including documents
  where optional fields are missing, before moving to production.

- Account for this behavior when setting confidence thresholds and assigning reviewers.

- If confidence scores for not-found fields are critical for your use case, select a model that uses
  calculated confidence scores.

## Model Errors

In some scenarios, the LLMs stop generating a response and provide an error code to understand the
reason that caused this behavior. If the error occurs during the execution of a document action, IDP
provides this error in the `statusMessage` attribute of the response object.

These are some of the most common error codes and their suggested fixes:

<table><colgroup><col> <col> <col></colgroup><thead><tr><th>Error Code</th><th>Error Description</th><th>Suggested Fix</th></tr></thead><tbody><tr><td><div><div><p><code>length</code></p></div></div></td><td><div><div><p>The request exceeded the maximum input token limit.</p></div></div></td><td><div><div><p>See <a href="https://developers.openai.com/api/docs/guides/text">Using long context</a> under GPT-4.1 prompting best practices to understand how to refine your prompts.</p></div></div></td></tr><tr><td><div><div><p><code>MAX_TOKENS</code></p></div></div></td><td><div><div><p>The response exceeded the maximum output token limit.</p></div></div></td><td><div><div><p>Split the document or select an OpenAI model for this extraction.</p></div></div></td></tr><tr><td><div><div><p><code>SPII</code></p></div></div></td><td><div><div><p>The document potentially contains Sensitive Personally Identifiable Information (SPII).</p></div></div></td><td><div><div><p>Delete the conflicting field, or use GPT models to abstract the information.</p></div></div></td></tr></tbody></table>

For a complete list of error codes, refer to each of the model provider’s documentation:

- OpenAI models:
  [The completion object](https://developers.openai.com/api/reference/resources/completions)

  Select **Show properties** under **choices** to see the details.

- Google models:
  [Finish Reason](https://docs.cloud.google.com/python/docs/reference/vertexai/latest?utm_source=chatgpt.com)

## See Also

- [OpenAI’s Prompt Engineering Guide](https://developers.openai.com/api/docs/guides/text)

- [Analyzing Documents With Custom User-Defined Schemas](https://docs.mulesoft.com/idp/analyzing-documents-with-einstein)