diff --git a/build_linux.sh b/build_linux.sh
index d072e2a..3b9571b 100755
--- a/build_linux.sh
+++ b/build_linux.sh
@@ -47,7 +47,8 @@ git clone https://github.com/Dao-AILab/flash-attention.git -b "v$FLASH_ATTN_VERS
 # Build wheels
 echo "Building wheels..."
 cd flash-attention
-FLASH_ATTENTION_FORCE_BUILD=TRUE python setup.py bdist_wheel --dist-dir=dist
+LOCAL_VERSION_LABEL="cu${MATRIX_CUDA_VERSION}torch${MATRIX_TORCH_VERSION}"
+FLASH_ATTENTION_FORCE_BUILD=TRUE FLASH_ATTN_LOCAL_VERSION=${LOCAL_VERSION_LABEL} python setup.py bdist_wheel --dist-dir=dist
 base_wheel_name=$(basename $(ls dist/*.whl | head -n 1))
 wheel_name=$(echo $base_wheel_name | sed "s/$FLASH_ATTN_VERSION/$FLASH_ATTN_VERSION+cu${MATRIX_CUDA_VERSION}torch${MATRIX_TORCH_VERSION}/")
 mv -v dist/$base_wheel_name dist/$wheel_name
diff --git a/build_windows.ps1 b/build_windows.ps1
index d67b5d6..c9f32e0 100644
--- a/build_windows.ps1
+++ b/build_windows.ps1
@@ -68,6 +68,7 @@ $env:FLASH_ATTENTION_FORCE_BUILD = "TRUE"
 $env:NVCC_FLAGS = "-w --disable-warnings"
 $env:CXXFLAGS = "/w"
 $env:CFLAGS = "/w"
+$env:FLASH_ATTN_LOCAL_VERSION = "cu$MatrixCudaVersion" + "torch$MatrixTorchVersion"
 
 cd flash-attention
 python setup.py bdist_wheel --dist-dir=dist
diff --git a/create_matrix.py b/create_matrix.py
index 831a6e7..62d2108 100644
--- a/create_matrix.py
+++ b/create_matrix.py
@@ -23,18 +23,21 @@ EXCLUDE = [
 
 LINUX_MATRIX = {
     "flash-attn-version": [
-        # "2.6.3", "2.7.4.post1", "2.8.3"
-        "2.8.1"
+        # "2.6.3", "2.7.4.post1"
+        "2.8.3"
+    ],
+    "python-version": [
+        # "3.10", "3.11", "3.12",
+        "3.13"
     ],
-    "python-version": ["3.10", "3.11", "3.12", "3.13"],
     "torch-version": [
         # "2.5.1", "2.6.0", "2.7.1", "2.8.0",
         "2.9.0",
     ],
     "cuda-version": [
         # "12.4.1", "12.6.3", "12.8.1", "12.9.1",
-        "12.8.1",
-        "13.0.1",
+        # "12.8.1",
+        "13.0.2",
     ],
 }
 
@@ -46,10 +49,22 @@ LINUX_SELF_HOSTED_MATRIX = {
 }
 
 WINDOWS_MATRIX = {
-    "flash-attn-version": ["2.7.4.post1", "2.8.3"],
-    "python-version": ["3.10", "3.11", "3.12", "3.13"],
-    "torch-version": ["2.5.1", "2.6.0", "2.7.1", "2.8.0", "2.9.0"],
-    "cuda-version": ["12.4.1", "12.6.3", "12.8.1", "12.9.1", "13.0.1"],
+    "flash-attn-version": [
+        # "2.7.4.post1",
+        "2.8.3"
+    ],
+    "python-version": [
+        # "3.10", "3.11", "3.12",
+        "3.13"
+    ],
+    "torch-version": [
+        # "2.5.1", "2.6.0", "2.7.1", "2.8.0",
+        "2.9.0"
+    ],
+    "cuda-version": [
+        # "12.4.1", "12.6.3", "12.8.1", "12.9.1",
+        "13.0.1"
+    ],
 }
 
 WINDOWS_CODEBUILD_MATRIX = {
@@ -68,8 +83,8 @@ def main():
                 # "linux": False,
                 # "linux_self_hosted": LINUX_SELF_HOSTED_MATRIX,
                 "linux_self_hosted": False,
-                # "windows": WINDOWS_MATRIX,
-                "windows": False,
+                "windows": WINDOWS_MATRIX,
+                # "windows": False,
                 # "windows_code_build": WINDOWS_CODEBUILD_MATRIX,
                 "windows_code_build": False,
                 "exclude": EXCLUDE,