diff --git a/third_party/cutlass b/third_party/cutlass
index 44c704eae85da..a75b4ac483166 160000
--- a/third_party/cutlass
+++ b/third_party/cutlass
@@ -1 +1 @@
-Subproject commit 44c704eae85da352d277d6f092f41412772f70e4
+Subproject commit a75b4ac483166189a45290783cb0a18af5ff0ea5
diff --git a/torch/_inductor/codegen/cuda/cutlass_lib_extensions/gemm_operation_extensions.py b/torch/_inductor/codegen/cuda/cutlass_lib_extensions/gemm_operation_extensions.py
index 16439f2520037..f7f32131d7bc4 100644
--- a/torch/_inductor/codegen/cuda/cutlass_lib_extensions/gemm_operation_extensions.py
+++ b/torch/_inductor/codegen/cuda/cutlass_lib_extensions/gemm_operation_extensions.py
@@ -52,7 +52,7 @@ def __init__(self, operation_suffix=""):
         >;
 
         using ADDMM_EVT =  // alpha * acc + beta * C
-            cutlass::epilogue::fusion::Sm90EVT<cutlass::epilogue::fusion::Sm90Compute<cutlass::multiply_add,
+            cutlass::epilogue::fusion::Sm90EVT<cutlass::epilogue::fusion::Sm90Compute<cutlass::homogeneous_multiply_add,
                     ElementD, ElementAcc, RoundStyle>, // beta * C + (alpha * acc)
               cutlass::epilogue::fusion::Sm90ScalarBroadcast<ElementAcc>, // beta
               cutlass::epilogue::fusion::Sm90SrcFetch, // C
diff --git a/torch/_inductor/codegen/cuda/cutlass_utils.py b/torch/_inductor/codegen/cuda/cutlass_utils.py
index 3835b3a5d6dbe..87fbf0008e1af 100644
--- a/torch/_inductor/codegen/cuda/cutlass_utils.py
+++ b/torch/_inductor/codegen/cuda/cutlass_utils.py
@@ -141,11 +141,9 @@ def _gen_ops_cached(arch, version) -> List[Any]:
 
     # Import cutlass python scripts.
     assert try_import_cutlass()
-    import torch._inductor.codegen.cuda.cutlass_lib_extensions.generator_extended_v322 as cutlass_generator  # type: ignore[import]
+    import torch._inductor.codegen.cuda.cutlass_lib_extensions.generator_extended as cutlass_generator  # type: ignore[import]
     import cutlass_library.manifest as cutlass_manifest  # type: ignore[import]
 
-    import torch._inductor.codegen.cuda.cutlass_lib_extensions.generator_extended_v322 as cutlass_generator  # type: ignore[import]
-
     if arch is None or version is None:
         log.error(
             "Cannot detect cuda arch %s or cuda version %s. "