diff --git a/ports/directml/portfile.cmake b/ports/directml/portfile.cmake
index 283ddaf1..b680db4c 100644
--- a/ports/directml/portfile.cmake
+++ b/ports/directml/portfile.cmake
@@ -5,7 +5,7 @@ set(ENV{NUGET_PACKAGES} "${BUILDTREES_DIR}/nuget")
 
 # see https://www.nuget.org/packages/Microsoft.AI.DirectML/
 set(PACKAGE_NAME    "Microsoft.AI.DirectML")
-set(PACKAGE_VERSION "1.13.0")
+set(PACKAGE_VERSION "1.13.1")
 
 file(REMOVE_RECURSE "${CURRENT_BUILDTREES_DIR}/${PACKAGE_NAME}")
 vcpkg_execute_required_process(
@@ -17,11 +17,16 @@ vcpkg_execute_required_process(
 
 get_filename_component(SOURCE_PATH "${CURRENT_BUILDTREES_DIR}/${PACKAGE_NAME}.${PACKAGE_VERSION}" ABSOLUTE)
 if(VCPKG_TARGET_IS_WINDOWS)
-    # todo: x64-xbox?
     if(VCPKG_TARGET_ARCHITECTURE STREQUAL "x64")
         set(TRIPLE "x64-win")
+        # check community triplets...
+        if(DEFINED VCPKG_XBOX_CONSOLE_TARGET AND (VCPKG_XBOX_CONSOLE_TARGET MATCHES "scarlett"))
+            set(TRIPLE "x64-xbox-scarlett-231000")
+        endif()
     elseif(VCPKG_TARGET_ARCHITECTURE STREQUAL "x86")
         set(TRIPLE "x86-win")
+    elseif(VCPKG_TARGET_ARCHITECTURE STREQUAL "arm64ec")
+        set(TRIPLE "arm64ec-win")
     elseif(VCPKG_TARGET_ARCHITECTURE STREQUAL "arm64")
         set(TRIPLE "arm64-win")
     elseif(VCPKG_TARGET_ARCHITECTURE STREQUAL "arm")
@@ -45,9 +50,6 @@ if(VCPKG_TARGET_IS_WINDOWS)
     file(INSTALL "${SOURCE_PATH}/bin/${TRIPLE}/DirectML.pdb"        DESTINATION "${CURRENT_PACKAGES_DIR}/bin")
     file(INSTALL "${SOURCE_PATH}/bin/${TRIPLE}/DirectML.Debug.dll"  DESTINATION "${CURRENT_PACKAGES_DIR}/bin")
     file(INSTALL "${SOURCE_PATH}/bin/${TRIPLE}/DirectML.Debug.pdb"  DESTINATION "${CURRENT_PACKAGES_DIR}/bin")
-    # debug/release will use same copy
-    file(INSTALL "${CURRENT_PACKAGES_DIR}/bin/" DESTINATION "${CURRENT_PACKAGES_DIR}/debug/bin")
-    file(INSTALL "${CURRENT_PACKAGES_DIR}/lib/" DESTINATION "${CURRENT_PACKAGES_DIR}/debug/lib")
 elseif(VCPKG_TARGET_IS_LINUX)
     file(INSTALL "${SOURCE_PATH}/bin/${TRIPLE}/libdirectml.so"  DESTINATION "${CURRENT_PACKAGES_DIR}/lib")
 else()
@@ -55,10 +57,11 @@ else()
 endif()
 
 file(INSTALL "${SOURCE_PATH}/include/" DESTINATION "${CURRENT_PACKAGES_DIR}/include")
+file(REMOVE_RECURSE "${CURRENT_PACKAGES_DIR}/debug")
 
-file(INSTALL "${SOURCE_PATH}/LICENSE.txt" DESTINATION "${CURRENT_PACKAGES_DIR}/share/${PORT}" RENAME copyright)
 file(INSTALL "${SOURCE_PATH}/LICENSE-CODE.txt"
              "${SOURCE_PATH}/README.md"
              "${SOURCE_PATH}/ThirdPartyNotices.txt"
      DESTINATION "${CURRENT_PACKAGES_DIR}/share/${PORT}"
 )
+vcpkg_install_copyright(FILE_LIST "${SOURCE_PATH}/LICENSE.txt")
diff --git a/ports/directml/vcpkg.json b/ports/directml/vcpkg.json
index 5d3df4a6..d1348c4d 100644
--- a/ports/directml/vcpkg.json
+++ b/ports/directml/vcpkg.json
@@ -1,6 +1,6 @@
 {
   "name": "directml",
-  "version-semver": "1.13.0",
+  "version-semver": "1.13.1",
   "description": [
     "DirectML is a high-performance, hardware-accelerated DirectX 12 library for machine learning.",
     "DirectML provides GPU acceleration for common machine learning tasks across a broad range of supported hardware and drivers, ",
diff --git a/ports/nvidia-cutlass/fix-cmake.patch b/ports/nvidia-cutlass/fix-cmake.patch
new file mode 100644
index 00000000..04544c4a
--- /dev/null
+++ b/ports/nvidia-cutlass/fix-cmake.patch
@@ -0,0 +1,50 @@
+diff --git a/CMakeLists.txt b/CMakeLists.txt
+index 114d793..7a05453 100644
+--- a/CMakeLists.txt
++++ b/CMakeLists.txt
+@@ -668,7 +668,7 @@ include(CTest)
+ enable_testing()
+ 
+ if (CUTLASS_ENABLE_GTEST_UNIT_TESTS)
+-  include(${CMAKE_CURRENT_SOURCE_DIR}/cmake/googletest.cmake)
++  find_package(GTest CONFIG REQUIRED) # GTest::gtest GTest::gmock
+ endif()
+ 
+ if (NOT TARGET test_all)
+diff --git a/test/unit/CMakeLists.txt b/test/unit/CMakeLists.txt
+index 16252a0..800ffc6 100644
+--- a/test/unit/CMakeLists.txt
++++ b/test/unit/CMakeLists.txt
+@@ -42,7 +42,7 @@ target_link_libraries(
+   CUTLASS
+   cutlass_tools_util_includes
+   $<$<BOOL:${CUTLASS_ENABLE_CUBLAS}>:nvidia::cublas>
+-  gtest
++  GTest::gtest
+   cudart
+   cuda_driver
+   )
+@@ -84,7 +84,7 @@ function(cutlass_test_unit_add_executable NAME)
+     target_link_libraries(
+       ${NAME}
+       PUBLIC
+-      gtest
++      GTest::gtest
+     )
+   else()
+     target_link_libraries(
+diff --git a/test/unit/core/CMakeLists.txt b/test/unit/core/CMakeLists.txt
+index 965c050..a11c0e8 100644
+--- a/test/unit/core/CMakeLists.txt
++++ b/test/unit/core/CMakeLists.txt
+@@ -45,6 +45,10 @@ cutlass_test_unit_add_executable(
+   functional.cu
+   )
+ 
++target_include_directories(cutlass_test_unit_core PUBLIC
++  $<BUILD_INTERFACE:${CUTLASS_INCLUDE_DIR}>
++  $<BUILD_INTERFACE:${CUTLASS_TOOLS_UTIL_INCLUDE_DIR}>
++)
+ #
+ # CUTLASS 3x increases the host compiler requirements to C++17. However, there are
+ # certain existing integrations that will benefit from maintaining C++11 compatibility.
diff --git a/ports/nvidia-cutlass/portfile.cmake b/ports/nvidia-cutlass/portfile.cmake
new file mode 100644
index 00000000..8fd7d036
--- /dev/null
+++ b/ports/nvidia-cutlass/portfile.cmake
@@ -0,0 +1,60 @@
+if(VCPKG_TARGET_IS_WINDOWS)
+    vcpkg_check_linkage(ONLY_STATIC_LIBRARY)
+endif()
+
+vcpkg_from_github(
+    OUT_SOURCE_PATH SOURCE_PATH
+    REPO NVIDIA/cutlass
+    REF v3.4.0
+    SHA512 6debdfa0ffbb9c14293619a281e864ac16592f64856e4a54ae456a79cbce3ec2f5068031f4eae589e72fb10c3c61f1fa6d3fa14650754a27c4b12e14774e7f75
+    PATCHES
+        fix-cmake.patch
+    HEAD_REF main
+)
+
+vcpkg_check_features(OUT_FEATURE_OPTIONS FEATURE_OPTIONS
+    FEATURES
+        tools CUTLASS_ENABLE_TOOLS
+        tools CUTLASS_ENABLE_PROFILER
+        test  CUTLASS_ENABLE_TESTS
+        test  CUTLASS_ENABLE_GTEST_UNIT_TESTS
+        test  CUTLASS_INSTALL_TESTS
+)
+
+if("tools" IN_LIST FEATURES)
+    x_vcpkg_get_python_packages(
+        PYTHON_VERSION 3
+        PACKAGES numpy
+        OUT_PYTHON_VAR PYTHON3
+    )
+    get_filename_component(PYTHON_PATH "${PYTHON3}" PATH)
+else()
+    vcpkg_find_acquire_program(PYTHON3)
+    get_filename_component(PYTHON_PATH "${PYTHON3}" PATH)
+endif()
+vcpkg_add_to_path(PREPEND "${PYTHON_PATH}")
+
+vcpkg_cmake_configure(
+    SOURCE_PATH "${SOURCE_PATH}"
+    OPTIONS
+        ${FEATURE_OPTIONS}
+        -DCUTLASS_REVISION:STRING=v3.4.0
+        -DCUTLASS_ENABLE_HEADERS_ONLY=ON
+        -DCUTLASS_ENABLE_LIBRARY=ON
+        -DCUTLASS_ENABLE_EXAMPLES=OFF
+        -DCUTLASS_ENABLE_PERFORMANCE=OFF
+        -DCUTLASS_LIBRARY_OPERATIONS:STRING=all
+        -DCUTLASS_LIBRARY_KERNELS:STRING=all
+        -DPython3_EXECUTABLE:FILEPATH=${PYTHON3}
+    MAYBE_UNUSED_VARIABLES
+        CUTLASS_LIBRARY_OPERATIONS
+)
+vcpkg_cmake_install()
+vcpkg_cmake_config_fixup(CONFIG_PATH lib/cmake/NvidiaCutlass PACKAGE_NAME NvidiaCutlass)
+
+file(REMOVE_RECURSE
+    "${CURRENT_PACKAGES_DIR}/debug"
+    "${CURRENT_PACKAGES_DIR}/test"
+    "${CURRENT_PACKAGES_DIR}/lib"
+)
+vcpkg_install_copyright(FILE_LIST "${SOURCE_PATH}/LICENSE.txt")
diff --git a/ports/nvidia-cutlass/vcpkg.json b/ports/nvidia-cutlass/vcpkg.json
new file mode 100644
index 00000000..9b5eee77
--- /dev/null
+++ b/ports/nvidia-cutlass/vcpkg.json
@@ -0,0 +1,18 @@
+{
+  "name": "nvidia-cutlass",
+  "version-semver": "3.4.0",
+  "description": "CUDA Templates for Linear Algebra Subroutines",
+  "homepage": "https://github.com/NVIDIA/cutlass",
+  "license": "BSD-3-Clause",
+  "dependencies": [
+    "cuda",
+    {
+      "name": "vcpkg-cmake",
+      "host": true
+    },
+    {
+      "name": "vcpkg-cmake-config",
+      "host": true
+    }
+  ]
+}
diff --git a/test/azure-port-arm64-windows.txt b/test/azure-port-arm64-windows.txt
index 5df976da..b7298152 100644
--- a/test/azure-port-arm64-windows.txt
+++ b/test/azure-port-arm64-windows.txt
@@ -1,2 +1,3 @@
 libdispatch
 tensorflow-lite
+onnxruntime[directml]
diff --git a/test/self-hosted-cuda.json b/test/self-hosted-cuda.json
index 0210bceb..f9875958 100644
--- a/test/self-hosted-cuda.json
+++ b/test/self-hosted-cuda.json
@@ -14,6 +14,7 @@
       "platform": "x64 & windows"
     },
     "nvidia-cnmem",
+    "nvidia-cutlass",
     "nvidia-tools-extension-sdk",
     "nvidia-triton-common",
     "nvidia-triton-core",
diff --git a/versions/baseline.json b/versions/baseline.json
index 58fbf023..bc81cb8e 100644
--- a/versions/baseline.json
+++ b/versions/baseline.json
@@ -25,7 +25,7 @@
       "port-version": 0
     },
     "directml": {
-      "baseline": "1.13.0",
+      "baseline": "1.13.1",
       "port-version": 0
     },
     "dlpack": {
@@ -92,6 +92,10 @@
       "baseline": "2018-11-28",
       "port-version": 0
     },
+    "nvidia-cutlass": {
+      "baseline": "3.4.0",
+      "port-version": 0
+    },
     "nvidia-tools-extension-sdk": {
       "baseline": "3.1.0",
       "port-version": 0
diff --git a/versions/d-/directml.json b/versions/d-/directml.json
index d9ba6e5e..59c8f49b 100644
--- a/versions/d-/directml.json
+++ b/versions/d-/directml.json
@@ -1,5 +1,10 @@
 {
   "versions": [
+    {
+      "git-tree": "aa4bc85c9d05b6aecb312b23f26f9d3c5014f5c8",
+      "version-semver": "1.13.1",
+      "port-version": 0
+    },
     {
       "git-tree": "b8621fdac36cd2db3cea051d27b8b2e8eb8dfb33",
       "version-semver": "1.13.0",
diff --git a/versions/n-/nvidia-cutlass.json b/versions/n-/nvidia-cutlass.json
new file mode 100644
index 00000000..d9c50cb3
--- /dev/null
+++ b/versions/n-/nvidia-cutlass.json
@@ -0,0 +1,9 @@
+{
+  "versions": [
+    {
+      "git-tree": "75c26a2bd20f4cbd49b9ed5189b9eb2a21e30007",
+      "version-semver": "3.4.0",
+      "port-version": 0
+    }
+  ]
+}