diff --git a/.github/workflows/test-wheel-windows.yml b/.github/workflows/test-wheel-windows.yml
index 478826c525..9022be4def 100644
--- a/.github/workflows/test-wheel-windows.yml
+++ b/.github/workflows/test-wheel-windows.yml
@@ -263,6 +263,7 @@ jobs:
           }
 
       - name: Install cuda.pathfinder extra wheels for testing
+        if: ${{ !endsWith(matrix.PY_VER, 't') }} # see issue #1820
         shell: bash --noprofile --norc -xeuo pipefail {0}
         run: |
           pushd cuda_pathfinder
@@ -271,6 +272,7 @@ jobs:
           popd
 
       - name: Run cuda.pathfinder tests with all_must_work
+        if: ${{ !endsWith(matrix.PY_VER, 't') }} # see issue #1820
         env:
           CUDA_PATHFINDER_TEST_LOAD_NVIDIA_DYNAMIC_LIB_STRICTNESS: all_must_work
           CUDA_PATHFINDER_TEST_FIND_NVIDIA_HEADERS_STRICTNESS: all_must_work
diff --git a/cuda_pathfinder/cuda/pathfinder/_dynamic_libs/descriptor_catalog.py b/cuda_pathfinder/cuda/pathfinder/_dynamic_libs/descriptor_catalog.py
index 69bb223a3d..e514b2e088 100644
--- a/cuda_pathfinder/cuda/pathfinder/_dynamic_libs/descriptor_catalog.py
+++ b/cuda_pathfinder/cuda/pathfinder/_dynamic_libs/descriptor_catalog.py
@@ -314,7 +314,7 @@ class DescriptorSpec:
         linux_sonames=("libmathdx.so.0",),
         windows_dlls=("mathdx64_0.dll",),
         site_packages_linux=("nvidia/cu13/lib", "nvidia/cu12/lib"),
-        site_packages_windows=("nvidia/cu13/bin/x86_64", "nvidia/cu12/bin"),
+        site_packages_windows=("nvidia/cu13/bin", "nvidia/cu12/bin"),
         dependencies=("nvrtc",),
     ),
     DescriptorSpec(
diff --git a/cuda_pathfinder/cuda/pathfinder/_headers/header_descriptor_catalog.py b/cuda_pathfinder/cuda/pathfinder/_headers/header_descriptor_catalog.py
index a46830e4ed..df1e52eb0f 100644
--- a/cuda_pathfinder/cuda/pathfinder/_headers/header_descriptor_catalog.py
+++ b/cuda_pathfinder/cuda/pathfinder/_headers/header_descriptor_catalog.py
@@ -145,6 +145,14 @@ class HeaderDescriptorSpec:
         conda_targets_layout=False,
         use_ctk_root_canary=False,
     ),
+    HeaderDescriptorSpec(
+        name="cute",
+        packaged_with="other",
+        header_basename="cute/tensor.hpp",
+        site_packages_dirs=("cutlass_library/source/include",),
+        conda_targets_layout=False,
+        use_ctk_root_canary=False,
+    ),
     HeaderDescriptorSpec(
         name="cutensor",
         packaged_with="other",
@@ -153,6 +161,22 @@ class HeaderDescriptorSpec:
         conda_targets_layout=False,
         use_ctk_root_canary=False,
     ),
+    HeaderDescriptorSpec(
+        name="cutlass",
+        packaged_with="other",
+        header_basename="cutlass/cutlass.h",
+        site_packages_dirs=("cutlass_library/source/include",),
+        conda_targets_layout=False,
+        use_ctk_root_canary=False,
+    ),
+    HeaderDescriptorSpec(
+        name="mathdx",
+        packaged_with="other",
+        header_basename="libmathdx.h",
+        site_packages_dirs=("nvidia/cu13/include", "nvidia/cu12/include"),
+        conda_targets_layout=False,
+        use_ctk_root_canary=False,
+    ),
     HeaderDescriptorSpec(
         name="nvshmem",
         packaged_with="other",
diff --git a/cuda_pathfinder/pyproject.toml b/cuda_pathfinder/pyproject.toml
index 872d30eb4d..2ccc253ac8 100644
--- a/cuda_pathfinder/pyproject.toml
+++ b/cuda_pathfinder/pyproject.toml
@@ -38,10 +38,12 @@ cu13 = [
     "nvidia-cudss-cu13",
     "nvidia-cufftmp-cu13; sys_platform != 'win32'",
     "nvidia-cusparselt-cu13",
+    "nvidia-libmathdx-cu13",
     "nvidia-nccl-cu13; sys_platform != 'win32'",
     "nvidia-nvshmem-cu13; sys_platform != 'win32'",
 ]
 host = [
+    "nvidia-cutlass",
     "nvpl-fft; platform_system == 'Linux' and platform_machine == 'aarch64'",
 ]
 
diff --git a/cuda_pathfinder/tests/test_find_nvidia_headers.py b/cuda_pathfinder/tests/test_find_nvidia_headers.py
index a47a235b2d..a7b95e167b 100644
--- a/cuda_pathfinder/tests/test_find_nvidia_headers.py
+++ b/cuda_pathfinder/tests/test_find_nvidia_headers.py
@@ -41,7 +41,10 @@
 
 NON_CTK_IMPORTLIB_METADATA_DISTRIBUTIONS_NAMES = {
     "cusparseLt": r"^nvidia-cusparselt-.*$",
+    "cute": r"^nvidia-cutlass$",
     "cutensor": r"^cutensor-.*$",
+    "cutlass": r"^nvidia-cutlass$",
+    "mathdx": r"^nvidia-libmathdx-.*$",
     "nvshmem": r"^nvidia-nvshmem-.*$",
 }
 
diff --git a/cuda_pathfinder/tests/test_load_nvidia_dynamic_lib.py b/cuda_pathfinder/tests/test_load_nvidia_dynamic_lib.py
index 016acfd25d..36487bd58e 100644
--- a/cuda_pathfinder/tests/test_load_nvidia_dynamic_lib.py
+++ b/cuda_pathfinder/tests/test_load_nvidia_dynamic_lib.py
@@ -90,7 +90,6 @@ def test_known_but_platform_unavailable_libname_raises_dynamic_lib_not_available
 
 
 IMPORTLIB_METADATA_DISTRIBUTIONS_NAMES = {
-    "cufftMp": r"^nvidia-cufftmp-.*$",
     "mathdx": r"^nvidia-libmathdx-.*$",
 }