Merge pull request #6408 from MicrosoftDocs/main

Auto Publish – main to live - 2025-08-06 05:06 UTC

Author: learn-build-service-prod[bot]

articles/ai-foundry/agents/how-to/virtual-networks.md

@@ -33,12 +33,13 @@ For customers without an existing virtual network, the Standard Setup with Priva
 :::image type="content" source="../media\private-network-isolation.png" alt-text="A diagram showing virtual network architecture.":::
 ### Known limitations
 
-- **Subnet IP address limitation**: both subnets must have IP ranges under `172.16.0.0/12` or `192.168.0.0/16`, i.e. class B or C address ranges reserved for private networking.
+- **Subnet IP address limitation**: both subnets must have IP ranges under `172.16.0.0/12` or `192.168.0.0/16`, i.e. class B or C private address ranges reserved for private networking. Public Class A, B or C address ranges are not supported. Private Class A address ranges have limited region support. For more information, see [our Private Network Secured Agent deployment template on GitHub](https://github.com/azure-ai-foundry/foundry-samples/blob/main/samples/microsoft/infrastructure-setup/15-private-network-standard-agent-setup/README.md).
 - **Agent subnet exclusivity**: The agent subnet cannot be shared by multiple Azure AI Foundry resources. Each AI Foundry must use a dedicated agent subnet.
+- **Agent subnet size**: The recommended size of the delegated Agent subnet is /24 (256 addresses) due to the delegation of the subnet to `Microsoft.App/environment`. For more on the subnet sizing, see [Configuring virtual networks for Azure Container Apps](/azure/container-apps/custom-virtual-networks?tabs=workload-profiles-env#subnet).
+- **Agent subnet egress firewall allowlisting**: If you are integrating an Azure Firewall with your private network secured standard agent, please allowlist the Fully Qualified Domain Names (FQDNs) listed in the [Integrate with Azure Firewall](/azure/container-apps/use-azure-firewall#application-rules) article.
 - **All Foundry workspace resources must be deployed in the same region as the virtual network (VNet)**. This includes Cosmos DB, Storage Account, AI Search, Foundry Account, Project, and Managed Identity.  
     - **Exception:** You may connect your Foundry Project to models deployed in a different region (on another AI Foundry or Azure OpenAI resource) by configuring an appropriate AI Services connection on the Project's [capability host](/azure/templates/microsoft.cognitiveservices/accounts/projects/capabilityhosts).
 - **Region availability**:
-  - For supported regions for Foundry workspace resources, see: [Azure AI Foundry project region availability](../../reference/region-support.md#azure-ai-foundry-projects).
   - For supported regions for model deployments, see: [Azure OpenAI model region support](../concepts/model-region-support.md#available-models).
 - **Azure Blob Storage**: using Azure Blob Storage files with the File Search tool isn't supported.
 

articles/ai-foundry/openai/concepts/models.md

@@ -22,6 +22,7 @@ Azure OpenAI is powered by a diverse set of models with different capabilities a
 
 | Models | Description |
 |--|--|
+| [gpt-oss](#gpt-oss) | **NEW** open-weight reasoning models |
 | [`codex-mini`](#o-series-models) | Fine-tuned version of o4-mini. |  
 | [`GPT-4.1 series`](#gpt-41-series) | The latest model release from Azure OpenAI. |
 | [`model-router`](#model-router) | A model that intelligently selects from a set of underlying chat models to respond to a given prompt. |
@@ -34,6 +35,38 @@ Azure OpenAI is powered by a diverse set of models with different capabilities a
 | [`Image generation`](#image-generation-models) | A series of models that can generate original images from natural language. |
 | [`Audio`](#audio-models) | A series of models for speech to text, translation, and text to speech. GPT-4o audio models support either low latency *speech in, speech out* conversational interactions or audio generation. |
 
+## gpt-oss
+
+### Region availability
+
+| Model | Region |
+|---|---|
+| `gpt-oss-120b`  | All Azure OpenAI regions |
+
+### Capabilities
+
+|  Model ID  | Description | Context Window | Max Output Tokens | Training Data (up to)  |
+|  --- |  :--- |:--- |:---|:---: |
+| `gpt-oss-120b` (Preview)   | - Text in/text out only <br> - Chat Completions API <br> - Streaming <br> - Function calling <br> - Structured outputs <br> - Reasoning <br> - Available for deployment<sup>1</sup> and via [managed compute](../../how-to/deploy-models-managed.md)  | 131,072 | 131,072 | May 31, 2024 |
+| `gpt-oss-20b` (Preview) | - Text in/text out only <br> - Chat Completions API <br> - Streaming <br> - Function calling <br> - Structured outputs <br> - Reasoning <br> - Available via [managed compute only](../../how-to/deploy-models-managed.md) | 131,072 | 131,072 | May 31, 2024 |
+
+<sup>1</sup> Unlike other Azure OpenAI models `gpt-oss-120b` requires an [Azure AI Foundry project](/azure/ai-foundry/quickstarts/get-started-code?tabs=azure-ai-foundry&pivots=fdp-project) to deploy the model.
+
+### Deploy with code
+
+```cli
+az cognitiveservices account deployment create \
+  --name "Foundry-project-resource" \
+  --resource-group "test-rg" \
+  --deployment-name "gpt-oss-120b" \
+  --model-name "gpt-oss-120b" \
+  --model-version "1" \
+  --model-format "OpenAI-OSS" \
+  --sku-capacity 10 \
+  --sku-name "GlobalStandard"
+```
+
+
 ## GPT-4.1 series
 
 ### Region availability

articles/ai-foundry/openai/quotas-limits.md

@@ -73,6 +73,12 @@ The following section provides you with a quick guide to the default quotas and
 
 [!INCLUDE [Quota](./includes/global-batch-limits.md)]
 
+## gpt-oss
+
+| Model          | Tokens per minute (TPM) | Requests per minute (RPM) |
+|----------------|-------------------|---------------------------------|
+| `gpt-oss-120b` | 5 M               | 5 K                             |
+
 ## GPT-4 rate limits
 
 ### GPT-4.5 preview Global Standard

articles/ai-services/content-understanding/concepts/classifier.md

@@ -1,5 +1,5 @@
 ---
-title: Azure AI Content Understanding classifier overview
+title: Azure AI Content Understanding Classifier Overview
 titleSuffix: Azure AI services
 description: Learn about Azure AI Content Understanding classifier solutions.
 author: PatrickFarley 
@@ -16,71 +16,68 @@ ms.custom:
 
 > [!IMPORTANT]
 >
-> * The classifier API is only available for documents with the `2025-05-01-preview` release.
-> * Azure AI Content Understanding classifier is available in `2025-05-01-preview` release. Public preview releases provide early access to features that are in active development.
-> * Features, approaches, and processes can change or have limited capabilities, before General Availability (GA).
-> * For more information, *see* [**Supplemental Terms of Use for Microsoft Azure Previews**](https://azure.microsoft.com/support/legal/preview-supplemental-terms).
+> The classifier API is available only for documents with the `2025-05-01-preview` release. The Azure AI Content Understanding classifier is available in the `2025-05-01-preview` release. Public preview releases provide early access to features that are in active development. Features, approaches, and processes can change or have limited capabilities before general availability. For more information, see [Supplemental Terms of Use for Microsoft Azure Previews](https://azure.microsoft.com/support/legal/preview-supplemental-terms).
 
-Azure AI Content Understanding classifier enables you to detect and identify documents you process within your application. Content Understanding classifier can perform classification of an input file as a whole, or identify multiple documents or multiple instances of a single document within an input file.
+You can use the Azure AI Content Understanding classifier to detect and identify documents that you process within your application. The Content Understanding classifier can perform classification of an input file as a whole. The classifier can also identify multiple documents or multiple instances of a single document within an input file.
 
 ## Business use cases
 
-Classifier can process complex documents in various formats and templates:
-
-* **Invoices**: Categorize invoices from multiple vendors to process each category with a different Content Understanding analyzer if needed.
-* **Tax documents**: Categorize multiple tax documents into different types of tax forms such as 1040, 1099, etc.
-* **Contracts**: Long, unstructured contracts can now be categorized to streamline operations to understand different types of agreements and their specific legal implications.
+The classifier can process complex documents in various formats and templates:
 
+* **Invoices**: Categorize invoices from multiple vendors to process each category with a different Content Understanding analyzer, if needed.
+* **Tax documents**: Categorize multiple tax documents into different types of tax forms, such as 1040 and 1099.
+* **Contracts**: Categorize long, unstructured contracts to streamline operations to understand different types of agreements and their specific legal implications.
 
 ## Content Understanding classifier capabilities
 
-Content Understanding classifier can analyze a single- or multi-file documents to identify if an input file can be classified into a category as defined. Here are the currently supported scenarios:
-
-* A single file containing one document type, such as a loan application form.
-* A single file containing multiple document types. For instance, a loan application package that contains a loan application form, payslip, and bank statement.
-* A single file containing multiple instances of the same document. For instance, a collection of scanned invoices.
-* By default, there's an `$OTHER` class as well, which we utilize for cases where any of the defined categories doesn't seem suitable.
+The Content Understanding classifier can analyze single or multifile documents to identify if an input file can be classified into a category as defined. The following scenarios are supported:
 
+* A single file that contains one document type, such as a loan application form.
+* A single file that contains multiple document types. An example is a loan application package that contains a loan application form, pay slip, and bank statement.
+* A single file that contains multiple instances of the same document. An example is a collection of scanned invoices.
+* By default, an `$OTHER` class is used for cases where none of the defined categories seems suitable.
 
-### How to use Content Understanding classifier
+### Use the Content Understanding classifier
 
-A Content Understanding classifier doesn't require any training dataset. Define up to 50 category name and description and create a classifier. By default, the entire file is treated as a single content object, meaning the file/object is associated to a single category.
+A Content Understanding classifier doesn't require any training dataset. You can define up to 50 category names and descriptions and create a classifier. By default, the entire file is treated as a single content object, which means the file or object is associated to a single category.
 
-However, when you have more than one document in a file, the classifier can identify the different document types contained within the input file with splitting capability. The classifier response contains the page ranges for each of the identified document types contained within a file. This response can include multiple instances of the same document type.
+When you have more than one document in a file, the classifier can identify the different document types that are contained within the input file with splitting capability. The classifier response contains the page ranges for each of the identified document types that are contained within a file. This response can include multiple instances of the same document type.
 
-When you call the classifier, the `analyze` operation includes a `splitMode` property that gives you granular control over the splitting behavior. You can also specify the page numbers to analyze only certain pages of the input document.
+When you call the classifier, the `analyze` operation includes a `splitMode` property that gives you granular control over the splitting behavior. You can also specify the page numbers to analyze only certain pages of the input document:
 
-* To treat the entire input file as a single document for classification set the `splitMode` to `none`. When you do so, the service returns just one category for the entire input file.
-* To classify each page of the input file, set the `splitMode` to `perPage`. The service attempts to classify each page as an individual document.
-* Set the `splitMode` to `auto` and the service identifies the documents and associated page ranges.
+* To treat the entire input file as a single document for classification, set `splitMode` to `none`. When you do so, the service returns one category for the entire input file.
+* To classify each page of the input file, set `splitMode` to `perPage`. The service attempts to classify each page as an individual document.
+* To identify the documents and associated page ranges, set `splitMode` to `auto`.
 
 ### Optional analysis
 
-For a complete end to end flow, you may link classifier categories with existing analyzers. For each content object classified to categories with linked analyzers, the service automatically invokes analysis on the content object using the corresponding analyzer. As an example, this linking can be used to create classifiers that identify and analyze only invoices from a PDF that may contain multiple types of forms in a document.
+For a complete end-to-end flow, you can link classifier categories with existing analyzers. For each content object classified to categories with linked analyzers, the service automatically invokes analysis on the content object by using the corresponding analyzer.
 
-* Set the `analyzerId` to an existing analyzer to route and perform field extraction from the classified documents or pages.
+For example, you can use this linking to create classifiers that identify and analyze only invoices from a PDF that contains multiple types of forms in a document. Set `analyzerId` to an existing analyzer to route and perform field extraction from the classified documents or pages.
 
 ### Classifier limits
 
-For information on supported input document formats and classifier limits, refer to our [Service quotas and limits](../service-limits.md#classifier) page.
-
+For information on supported input document formats and classifier limits, see [Service quotas and limits](../service-limits.md#classifier).
 
 ### Best practices
 
-To improve classification and splitting quality, it's important to give a good category name and description so the model can understand the categories with some context. For more information on category names and descriptions, *see* [Best practices](../concepts/best-practices.md#classifier-category-names-and-descriptions).
+To improve classification and splitting quality, use a good category name and description so that the model can understand the categories with some context. For more information on category names and descriptions, see [Best practices](../concepts/best-practices.md#classifier-category-names-and-descriptions).
 
 ## Key benefits
 
-* **Accuracy and reliability:** Ensure precise document classification, reducing errors and boosting efficiency.
-* **Scalability:** Seamlessly scale out document processing to meet business demands.
-* **Customizable:** Adapt document classifier to fit specific workflows.
+* **Accuracy and reliability**: Ensure precise document classification to reduce errors and boost efficiency.
+* **Scalability**: Scale out document processing to meet business demands.
+* **Customizable**: Adapt the document classifier to fit specific workflows.
 
 ## Supported languages and regions
-For a detailed list of supported languages and regions, visit our [Language and region support](../language-region-support.md) page.
+
+For a list of supported languages and regions, see [Language and region support](../language-region-support.md).
 
 ## Data privacy and security
-Developers using Content Understanding should review Microsoft's policies on customer data. For more information, visit our [Data, protection, and privacy](https://www.microsoft.com/trust-center/privacy) page.
 
-## Next step
-* Try processing your document content using Content Understanding in [Azure AI Foundry](https://aka.ms/cu-landing).
-* Learn to analyze document content [**analyzer templates**](../quickstart/use-ai-foundry.md).
+Developers who use Content Understanding should review Microsoft policies on customer data. For more information, see [Data, protection, and privacy](https://www.microsoft.com/trust-center/privacy).
+
+## Related content
+
+* Try processing your document content by using Content Understanding in [Azure AI Foundry](https://aka.ms/cu-landing).
+* Learn to analyze document content [analyzer templates](../quickstart/use-ai-foundry.md).