diff --git a/benchmarks/benchmark_aq.py b/benchmarks/benchmark_aq.py
index 09fe1bc452..fc19d1ab8a 100644
--- a/benchmarks/benchmark_aq.py
+++ b/benchmarks/benchmark_aq.py
@@ -5,7 +5,7 @@
     Int8WeightOnlyQuantizedLinearWeight,
     Int4WeightOnlyQuantizedLinearWeight,
 )
-from torchao.quantization.utils import (
+from torchao.utils import (
     TORCH_VERSION_AFTER_2_4,
 )
 from torchao.quantization.quant_api import (
diff --git a/benchmarks/intmm.py b/benchmarks/intmm.py
index edadf2a7cf..950f100f20 100644
--- a/benchmarks/intmm.py
+++ b/benchmarks/intmm.py
@@ -6,7 +6,7 @@
 import pathlib
 
 import torch
-from torchao.quantization.utils import TORCH_VERSION_AFTER_2_4, TORCH_VERSION_AFTER_2_2
+from torchao.utils import TORCH_VERSION_AFTER_2_4, TORCH_VERSION_AFTER_2_2
 
 
 # Check if CUDA is available, if not, exit the script