diff --git a/cmake/external/dml.cmake b/cmake/external/dml.cmake
index af1e5c940f..49a2d7a144 100644
--- a/cmake/external/dml.cmake
+++ b/cmake/external/dml.cmake
@@ -20,18 +20,18 @@ if (NOT onnxruntime_USE_CUSTOM_DIRECTML)
   set(NUGET_CONFIG ${PROJECT_SOURCE_DIR}/../NuGet.config)
   set(PACKAGES_CONFIG ${PROJECT_SOURCE_DIR}/../packages.config)
   get_filename_component(PACKAGES_DIR ${CMAKE_CURRENT_BINARY_DIR}/../packages ABSOLUTE)
-  set(DML_PACKAGE_DIR ${PACKAGES_DIR}/DirectML.1.3.0)
+  set(DML_PACKAGE_DIR ${PACKAGES_DIR}/Microsoft.AI.DirectML.1.4.0)
   set(DML_SHARED_LIB DirectML.dll)
 
   # Restore nuget packages, which will pull down the DirectML redist package
   add_custom_command(
-    OUTPUT ${DML_PACKAGE_DIR}/bin/x64/DirectML.lib ${DML_PACKAGE_DIR}/bin/x86/DirectML.lib
+    OUTPUT ${DML_PACKAGE_DIR}/bin/x64-win/DirectML.lib ${DML_PACKAGE_DIR}/bin/x86-win/DirectML.lib
     DEPENDS ${PACKAGES_CONFIG} ${NUGET_CONFIG}
     COMMAND ${CMAKE_CURRENT_BINARY_DIR}/nuget/src/nuget restore ${PACKAGES_CONFIG} -PackagesDirectory ${PACKAGES_DIR} -ConfigFile ${NUGET_CONFIG}
     VERBATIM)
 
   include_directories(BEFORE "${DML_PACKAGE_DIR}/include")
-  add_custom_target(RESTORE_PACKAGES ALL DEPENDS ${DML_PACKAGE_DIR}/bin/x64/DirectML.lib ${DML_PACKAGE_DIR}/bin/x86/DirectML.lib)
+  add_custom_target(RESTORE_PACKAGES ALL DEPENDS ${DML_PACKAGE_DIR}/bin/x64-win/DirectML.lib ${DML_PACKAGE_DIR}/bin/x86-win/DirectML.lib)
   add_dependencies(RESTORE_PACKAGES nuget)
 else()
   include_directories(${dml_INCLUDE_DIR})
diff --git a/cmake/onnxruntime_providers.cmake b/cmake/onnxruntime_providers.cmake
index 4692dfd4de..2afa476a44 100644
--- a/cmake/onnxruntime_providers.cmake
+++ b/cmake/onnxruntime_providers.cmake
@@ -776,7 +776,7 @@ if (onnxruntime_USE_DML)
       add_custom_command(TARGET onnxruntime_providers_dml
         POST_BUILD
         COMMAND ${CMAKE_COMMAND} -E copy_if_different
-          "${DML_PACKAGE_DIR}/bin/${onnxruntime_target_platform}/${file}" $<TARGET_FILE_DIR:onnxruntime_providers_dml>)
+          "${DML_PACKAGE_DIR}/bin/${onnxruntime_target_platform}-win/${file}" $<TARGET_FILE_DIR:onnxruntime_providers_dml>)
     endforeach()
   endif()
 
@@ -785,8 +785,8 @@ if (onnxruntime_USE_DML)
       target_link_libraries(${target} PRIVATE DirectML)
     else()
       add_dependencies(${target} RESTORE_PACKAGES)
-      target_link_libraries(${target} PRIVATE "${DML_PACKAGE_DIR}/bin/${onnxruntime_target_platform}/DirectML.lib")
-      target_compile_definitions(${target} PRIVATE DML_TARGET_VERSION_USE_LATEST)
+      target_link_libraries(${target} PRIVATE "${DML_PACKAGE_DIR}/bin/${onnxruntime_target_platform}-win/DirectML.lib")
+	    target_compile_definitions(${target} PRIVATE DML_TARGET_VERSION_USE_LATEST)
     endif()
   endfunction()
 
diff --git a/cmake/onnxruntime_python.cmake b/cmake/onnxruntime_python.cmake
index 81088e9b49..5fbfa3fdfb 100644
--- a/cmake/onnxruntime_python.cmake
+++ b/cmake/onnxruntime_python.cmake
@@ -396,7 +396,7 @@ if (onnxruntime_USE_DML)
   add_custom_command(
     TARGET onnxruntime_pybind11_state POST_BUILD
     COMMAND ${CMAKE_COMMAND} -E copy
-        ${DML_PACKAGE_DIR}/bin/${onnxruntime_target_platform}/${DML_SHARED_LIB}
+        ${DML_PACKAGE_DIR}/bin/${onnxruntime_target_platform}-win/${DML_SHARED_LIB}
         $<TARGET_FILE_DIR:${test_data_target}>/onnxruntime/capi/
   )
 endif()
diff --git a/csharp/src/Microsoft.AI.MachineLearning.Interop/Microsoft.AI.MachineLearning.targets b/csharp/src/Microsoft.AI.MachineLearning.Interop/Microsoft.AI.MachineLearning.targets
index 74a611d67f..68a53643b5 100644
--- a/csharp/src/Microsoft.AI.MachineLearning.Interop/Microsoft.AI.MachineLearning.targets
+++ b/csharp/src/Microsoft.AI.MachineLearning.Interop/Microsoft.AI.MachineLearning.targets
@@ -13,17 +13,6 @@
     <RuntimesDirectory Condition="'$(RuntimesDirectory)' == ''">$(MSBuildThisFileDirectory)..\..\runtimes\win-$(WindowsAI-Platform)\native</RuntimesDirectory>
     <WindowsAIBinary>$(RuntimesDirectory)\Microsoft.AI.MachineLearning.dll</WindowsAIBinary>
     <OnnxRuntimeBinary>$(RuntimesDirectory)\onnxruntime.dll</OnnxRuntimeBinary>
-    <DirectMLBinary>$(RuntimesDirectory)\directml.dll</DirectMLBinary>
-  </PropertyGroup>
-
-  <PropertyGroup>
-    <MLBinaries>$(WindowsAIBinary);$(OnnxRuntimeBinary);$(DirectMLBinary)</MLBinaries>
-  </PropertyGroup>
-  <PropertyGroup Condition="'$(WindowsAI-Platform)' == x64 OR '$(WindowsAI-Platform)' == x86">
-    <MLBinaries>$(WindowsAIBinary);$(OnnxRuntimeBinary);$(DirectMLBinary)</MLBinaries>
-  </PropertyGroup>
-  <PropertyGroup Condition="'$(OutputType)' == 'AppContainerExe' OR '$(WindowsAI-Platform)' == arm OR '$(WindowsAI-Platform)' == arm64">
-    <MLBinaries>$(WindowsAIBinary);$(OnnxRuntimeBinary)</MLBinaries>
   </PropertyGroup>
 
   <Target Name="CopyMLBinaries" BeforeTargets="CoreBuild">
@@ -31,6 +20,6 @@
       <WindowsAIBinplaceMessage>Binplacing WindowsAI binaries: {0} and {1}.</WindowsAIBinplaceMessage>
     </PropertyGroup>
     <Message Text="$([System.String]::Format('$(WindowsAIBinplaceMessage)', '$(WindowsAIBinary)', '$(OnnxRuntimeBinary)'))" />
-    <Copy SkipUnchangedFiles="True" SourceFiles="$(MLBinaries)" DestinationFolder="$(OutDir)" />
+    <Copy SkipUnchangedFiles="True" SourceFiles="$(WindowsAIBinary);$(OnnxRuntimeBinary)" DestinationFolder="$(OutDir)" />
   </Target>
 </Project>
diff --git a/csharp/src/Microsoft.AI.MachineLearning/Microsoft.AI.MachineLearning.Rules.Project.xml b/csharp/src/Microsoft.AI.MachineLearning/Microsoft.AI.MachineLearning.Rules.Project.xml
index 7492025563..6e1599f102 100644
--- a/csharp/src/Microsoft.AI.MachineLearning/Microsoft.AI.MachineLearning.Rules.Project.xml
+++ b/csharp/src/Microsoft.AI.MachineLearning/Microsoft.AI.MachineLearning.Rules.Project.xml
@@ -13,11 +13,6 @@
                 Description="Suppresses binplacing the onnxruntime.dll from the Mirosoft.AI.MachineLearning NuGet distribution."
                 Category="General" />
 
-  <BoolProperty Name="DisableDirectMLDllCopy"
-                DisplayName="Disable directml.dll copy on build"
-                Description="Suppresses binplacing the directml.dll from the Mirosoft.AI.MachineLearning NuGet distribution."
-                Category="General" />
-
   <BoolProperty Name="UseWindowsMLStaticRuntime"
                 DisplayName="Use a dll that links statically to runtimes"
                 Description="Use a larger DLL that doesn't depend on installed versions of the vcruntime and CRT."
diff --git a/csharp/src/Microsoft.AI.MachineLearning/Microsoft.AI.MachineLearning.props b/csharp/src/Microsoft.AI.MachineLearning/Microsoft.AI.MachineLearning.props
index 69acb49f1a..ec7bda8625 100644
--- a/csharp/src/Microsoft.AI.MachineLearning/Microsoft.AI.MachineLearning.props
+++ b/csharp/src/Microsoft.AI.MachineLearning/Microsoft.AI.MachineLearning.props
@@ -12,4 +12,11 @@
   <ItemGroup>
     <PropertyPageSchema Include="$(MSBuildThisFileDirectory)\Microsoft.AI.MachineLearning.Rules.Project.xml"/>
   </ItemGroup>
+
+  <PropertyGroup Label="Globals">
+    <!-- Assume apps using the WinML package only want the DirectML binaries (no need for a build dependency). -->
+    <Microsoft_AI_DirectML_SkipDebugLayerCopy>true</Microsoft_AI_DirectML_SkipDebugLayerCopy>
+    <Microsoft_AI_DirectML_SkipLink>true</Microsoft_AI_DirectML_SkipLink>
+    <Microsoft_AI_DirectML_SkipIncludeDir>true</Microsoft_AI_DirectML_SkipIncludeDir>
+  </PropertyGroup>
 </Project>
diff --git a/csharp/src/Microsoft.AI.MachineLearning/Microsoft.AI.MachineLearning.targets b/csharp/src/Microsoft.AI.MachineLearning/Microsoft.AI.MachineLearning.targets
index 4f43996fc9..bfecff8c98 100644
--- a/csharp/src/Microsoft.AI.MachineLearning/Microsoft.AI.MachineLearning.targets
+++ b/csharp/src/Microsoft.AI.MachineLearning/Microsoft.AI.MachineLearning.targets
@@ -15,25 +15,14 @@
   <PropertyGroup Condition="'$(DisableOnnxRuntimeDllCopy)' != 'true'">
     <OnnxRuntimeBinary Condition="'$(OnnxRuntimeBinary)' == ''">$(RuntimesDirectory)\$(WindowsAIBinarySubfolder)onnxruntime.dll</OnnxRuntimeBinary>
   </PropertyGroup>
-
-  <PropertyGroup Condition="'$(DisableDirectMLDllCopy)' != 'true'">
-    <DirectMLBinary Condition="'$(DirectMLBinary)' == ''">$(RuntimesDirectory)\directml.dll</DirectMLBinary>
-  </PropertyGroup>
   
   <ItemGroup>
     <Reference Include="$(MSBuildThisFileDirectory)\..\..\lib\uap\Microsoft.AI.MachineLearning.winmd">
       <Implementation>$(WindowsAIBinary)</Implementation>
     </Reference>
   </ItemGroup>
-  
-  <PropertyGroup Condition="'$(WindowsAI-Platform)' == x64 OR '$(WindowsAI-Platform)' == x86">
-    <MLBinaries>$(WindowsAIBinary);$(OnnxRuntimeBinary);$(DirectMLBinary)</MLBinaries>
-  </PropertyGroup>
-  <PropertyGroup Condition="'$(OutputType)' == 'AppContainerExe' OR '$(WindowsAI-Platform)' == arm OR '$(WindowsAI-Platform)' == arm64">
-    <MLBinaries>$(WindowsAIBinary);$(OnnxRuntimeBinary)</MLBinaries>
-  </PropertyGroup>
 
   <ItemGroup>
-    <ReferenceCopyLocalPaths Include="$(MLBinaries)" />
+    <ReferenceCopyLocalPaths Include="$(WindowsAIBinary);$(OnnxRuntimeBinary)" />
   </ItemGroup>
 </Project>
diff --git a/csharp/src/Microsoft.ML.OnnxRuntime/Microsoft.ML.OnnxRuntime.csproj b/csharp/src/Microsoft.ML.OnnxRuntime/Microsoft.ML.OnnxRuntime.csproj
index cf650200d0..803e191181 100644
--- a/csharp/src/Microsoft.ML.OnnxRuntime/Microsoft.ML.OnnxRuntime.csproj
+++ b/csharp/src/Microsoft.ML.OnnxRuntime/Microsoft.ML.OnnxRuntime.csproj
@@ -113,14 +113,6 @@
           CopyToOutputDirectory="Always"
           Visible="false"
     />
-    <None Include="$(NativeBuildOutputDir)\DirectML.dll"
-          Condition="Exists('$(NativeBuildOutputDir)\DirectML.dll')
-                     And '$(OrtPackageId)' == 'Microsoft.ML.OnnxRuntime.DirectML'"
-          PackagePath="\runtimes\win-$(TargetArchitecture)\native"
-          Pack="false"
-          CopyToOutputDirectory="Always"
-          Visible="false"
-    />
     <None Include="$(NativeBuildOutputDir)\dnnl.dll"
           Condition="Exists('$(NativeBuildOutputDir)\dnnl.dll')"
           PackagePath="\runtimes\win-$(TargetArchitecture)\native"
diff --git a/csharp/src/Microsoft.ML.OnnxRuntime/props.xml b/csharp/src/Microsoft.ML.OnnxRuntime/props.xml
index 1b763c3c4e..591b127518 100644
--- a/csharp/src/Microsoft.ML.OnnxRuntime/props.xml
+++ b/csharp/src/Microsoft.ML.OnnxRuntime/props.xml
@@ -29,7 +29,13 @@
 
   <PropertyGroup>
     <OnnxRuntimeBinary>$(MSBuildThisFileDirectory)..\..\runtimes\win-$(EnginePlatform)\native\onnxruntime.dll</OnnxRuntimeBinary>
-    <DirectMLBinary Condition="Exists('$(MSBuildThisFileDirectory)..\..\runtimes\win-$(EnginePlatform)\native\directml.dll')">$(MSBuildThisFileDirectory)..\..\runtimes\win-$(EnginePlatform)\native\directml.dll</DirectMLBinary>
+  </PropertyGroup>
+
+  <!-- Assume apps using the Microsoft.ML.OnnxRuntime.DirectML package only want the DirectML binaries (no need for a build dependency). -->
+  <PropertyGroup Label="Globals" Condition="Exists('$(MSBuildThisFileDirectory)include\dml_provider_factory.h')">
+    <Microsoft_AI_DirectML_SkipDebugLayerCopy>true</Microsoft_AI_DirectML_SkipDebugLayerCopy>
+    <Microsoft_AI_DirectML_SkipLink>true</Microsoft_AI_DirectML_SkipLink>
+    <Microsoft_AI_DirectML_SkipIncludeDir>true</Microsoft_AI_DirectML_SkipIncludeDir>
   </PropertyGroup>
 
   <ItemGroup>
diff --git a/docs/execution_providers/DirectML-ExecutionProvider.md b/docs/execution_providers/DirectML-ExecutionProvider.md
index f74fb45acd..3a1c7b5292 100644
--- a/docs/execution_providers/DirectML-ExecutionProvider.md
+++ b/docs/execution_providers/DirectML-ExecutionProvider.md
@@ -6,7 +6,7 @@ When used standalone, the DirectML API is a low-level DirectX 12 library and is
 
 The *DirectML Execution Provider* is an optional component of ONNX Runtime that uses DirectML to accelerate inference of ONNX models. The DirectML execution provider is capable of greatly improving evaluation time of models using commodity GPU hardware, without sacrificing broad hardware support or requiring vendor-specific extensions to be installed.
 
-The DirectML Execution Provider currently uses DirectML version 1.3.0.
+The DirectML Execution Provider currently uses DirectML version 1.4.0.
 
 ## Table of contents
 
@@ -15,11 +15,12 @@ The DirectML Execution Provider currently uses DirectML version 1.3.0.
   - [Minimum requirements](#minimum-requirements)
   - [Building from source](#building-from-source)
   - [Using the DirectML execution provider](#using-the-directml-execution-provider)
-    - [`OrtSessionOptionsAppendExecutionProvider_DML` function](#ortsessionoptionsappendexecutionproviderdml-function)
-    - [`OrtSessionOptionsAppendExecutionProviderEx_DML` function](#ortsessionoptionsappendexecutionproviderexdml-function)
+    - [`OrtSessionOptionsAppendExecutionProvider_DML` function](#ortsessionoptionsappendexecutionprovider_dml-function)
+    - [`OrtSessionOptionsAppendExecutionProviderEx_DML` function](#ortsessionoptionsappendexecutionproviderex_dml-function)
     - [ONNX opset support](#onnx-opset-support)
     - [Multi-threading and supported session options](#multi-threading-and-supported-session-options)
   - [Samples](#samples)
+  - [Performance best practices](#performance-best-practices)
   - [See also](#see-also)
 
 ## Minimum requirements
diff --git a/packages.config b/packages.config
index 6fce6d078a..cf2b71e61a 100644
--- a/packages.config
+++ b/packages.config
@@ -1,5 +1,5 @@
 <?xml version="1.0" encoding="utf-8"?>
 <packages>
-  <package id="DirectML" version="1.3.0" targetFramework="native" />
+  <package id="Microsoft.AI.DirectML" version="1.4.0" targetFramework="native" />
   <package id="GoogleTestAdapter" version="0.17.1" targetFramework="net46" />
 </packages>
diff --git a/tools/ci_build/github/azure-pipelines/nuget/templates/gpu.yml b/tools/ci_build/github/azure-pipelines/nuget/templates/gpu.yml
index 5ade2044cb..d694d2c846 100644
--- a/tools/ci_build/github/azure-pipelines/nuget/templates/gpu.yml
+++ b/tools/ci_build/github/azure-pipelines/nuget/templates/gpu.yml
@@ -200,8 +200,6 @@ jobs:
              move win-x86\runtimes\win-x86\native\onnxruntime.dll %%~ni\runtimes\win-x86\native\onnxruntime.dll
              move win-x86\runtimes\win-x86\native\onnxruntime.lib %%~ni\runtimes\win-x86\native\onnxruntime.lib
              move win-x86\runtimes\win-x86\native\onnxruntime.pdb %%~ni\runtimes\win-x86\native\onnxruntime.pdb
-             move win-x86\runtimes\win-x86\native\directml.dll %%~ni\runtimes\win-x86\native\DirectML.dll
-             move win-x86\runtimes\win-x86\native\directml.pdb %%~ni\runtimes\win-x86\native\DirectML.pdb
 
              pushd %%~ni
              zip -r ..\%%~ni.zip .
diff --git a/tools/ci_build/github/azure-pipelines/nuget/templates/windowsai.yml b/tools/ci_build/github/azure-pipelines/nuget/templates/windowsai.yml
index 8ff5811300..417138d94d 100644
--- a/tools/ci_build/github/azure-pipelines/nuget/templates/windowsai.yml
+++ b/tools/ci_build/github/azure-pipelines/nuget/templates/windowsai.yml
@@ -363,7 +363,6 @@ jobs:
         Copy-Item ([System.IO.Path]::Combine($x86_runtime_path_old, 'onnxruntime.lib'))                    $x86_runtime_path_new
         Copy-Item ([System.IO.Path]::Combine($x86_runtime_path_old, 'microsoft.ai.machinelearning.dll'))   $x86_runtime_path_new
         Copy-Item ([System.IO.Path]::Combine($x86_runtime_path_old, 'microsoft.ai.machinelearning.lib'))   $x86_runtime_path_new
-        Copy-Item ([System.IO.Path]::Combine($x86_runtime_path_old, 'directml.dll'))                       $x86_runtime_path_new
 
         Copy-Item ([System.IO.Path]::Combine($arm64_runtime_path_old, 'onnxruntime.dll'))                  $arm64_runtime_path_new
         Copy-Item ([System.IO.Path]::Combine($arm64_runtime_path_old, 'onnxruntime.lib'))                  $arm64_runtime_path_new
diff --git a/tools/nuget/generate_nuspec_for_native_nuget.py b/tools/nuget/generate_nuspec_for_native_nuget.py
index aa74ccb82a..75b4e3d44c 100644
--- a/tools/nuget/generate_nuspec_for_native_nuget.py
+++ b/tools/nuget/generate_nuspec_for_native_nuget.py
@@ -76,6 +76,8 @@ def generate_repo_url(list, repo_url, commit_id):
 
 
 def generate_dependencies(list, package_name, version):
+    dml_dependency = '<dependency id="Microsoft.AI.DirectML" version="1.4.0"/>'
+
     if (package_name == 'Microsoft.AI.MachineLearning'):
         list.append('<dependencies>')
 
@@ -96,23 +98,41 @@ def generate_dependencies(list, package_name, version):
         list.append('</group>')
         # UAP10.0.16299, This is the earliest release of the OS that supports .NET Standard apps
         list.append('<group targetFramework="UAP10.0.16299">')
+        list.append(dml_dependency)
+        list.append('</group>')
+        # Support Native C++
+        list.append('<group targetFramework="native">')
+        list.append(dml_dependency)
         list.append('</group>')
 
         list.append('</dependencies>')
     else:
+        include_dml = package_name == 'Microsoft.ML.OnnxRuntime.DirectML'
+
         list.append('<dependencies>')
         # Support .Net Core
         list.append('<group targetFramework="NETCOREAPP">')
         list.append('<dependency id="Microsoft.ML.OnnxRuntime.Managed"' + ' version="' + version + '"/>')
+        if include_dml:
+            list.append(dml_dependency)
         list.append('</group>')
         # Support .Net Standard
         list.append('<group targetFramework="NETSTANDARD">')
         list.append('<dependency id="Microsoft.ML.OnnxRuntime.Managed"' + ' version="' + version + '"/>')
+        if include_dml:
+            list.append(dml_dependency)
         list.append('</group>')
         # Support .Net Framework
         list.append('<group targetFramework="NETFRAMEWORK">')
         list.append('<dependency id="Microsoft.ML.OnnxRuntime.Managed"' + ' version="' + version + '"/>')
+        if include_dml:
+            list.append(dml_dependency)
         list.append('</group>')
+        # Support Native C++
+        if include_dml:
+            list.append('<group targetFramework="native">')
+            list.append(dml_dependency)
+            list.append('</group>')
 
         list.append('</dependencies>')
 
@@ -289,14 +309,6 @@ def generate_files(list, args):
                           'libonnxruntime.so') + '" target="runtimes\\linux-' + args.target_architecture +
                           '\\native" />')
 
-    if includes_directml:
-        files_list.append('<file src=' + '"' + os.path.join(args.native_build_path, 'DirectML.dll') +
-                          runtimes + ' />')
-        files_list.append('<file src=' + '"' + os.path.join(args.native_build_path, 'DirectML.pdb') +
-                          runtimes + ' />')
-        files_list.append('<file src=' + '"' + os.path.join(args.packages_path, 'DirectML.1.3.0\\LICENSE.txt') +
-                          '" target="DirectML_LICENSE.txt" />')
-
     if includes_winml:
         # Process microsoft.ai.machinelearning import lib, dll, and pdb
         files_list.append('<file src=' + '"' +