diff --git a/cmake/CMakeLists.txt b/cmake/CMakeLists.txt
index 5f4c104009f53..a35671c461fba 100644
--- a/cmake/CMakeLists.txt
+++ b/cmake/CMakeLists.txt
@@ -285,6 +285,7 @@ set_target_properties(onnx_proto PROPERTIES FOLDER "External/ONNX")
 # fix a warning in onnx code we can't do anything about
 if (MSVC)
     target_compile_options(onnx_proto PRIVATE /wd4146) # unary minus operator applied to unsigned type
+    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -DEIGEN_HAS_C99_MATH") # required to be set explicitly to enable Eigen-Unsupported SpecialFunctions
 endif()
 set(onnxruntime_EXTERNAL_DEPENDENCIES gsl onnx_proto)
 
diff --git a/onnxruntime/core/providers/cpu/cpu_execution_provider.cc b/onnxruntime/core/providers/cpu/cpu_execution_provider.cc
index f5303df02c82f..cdcb09e523654 100644
--- a/onnxruntime/core/providers/cpu/cpu_execution_provider.cc
+++ b/onnxruntime/core/providers/cpu/cpu_execution_provider.cc
@@ -195,6 +195,7 @@ class ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, Con
 class ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, EyeLike);
 class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, float, IsNaN);
 class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, MLFloat16, IsNaN);
+class ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, Erf);
 
 void RegisterOnnxOperatorKernels(std::function<void(KernelCreateInfo&&)> fn) {
   fn(BuildKernel<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, Clip)>());
@@ -382,6 +383,7 @@ void RegisterOnnxOperatorKernels(std::function<void(KernelCreateInfo&&)> fn) {
   fn(BuildKernel<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, EyeLike)>());
   fn(BuildKernel<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, float, IsNaN)>());
   fn(BuildKernel<ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, MLFloat16, IsNaN)>());
+  fn(BuildKernel<ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, Erf)>());
 }
 
 // Forward declarations of ml op kernels
diff --git a/onnxruntime/core/providers/cpu/math/element_wise_ops.cc b/onnxruntime/core/providers/cpu/math/element_wise_ops.cc
index 8f48195ea0145..ab05ed82be80c 100644
--- a/onnxruntime/core/providers/cpu/math/element_wise_ops.cc
+++ b/onnxruntime/core/providers/cpu/math/element_wise_ops.cc
@@ -2,6 +2,7 @@
 // Licensed under the MIT License.
 
 #include "core/providers/cpu/math/element_wise_ops.h"
+#include <unsupported/Eigen/SpecialFunctions>
 
 namespace onnxruntime {
 
@@ -311,6 +312,12 @@ ONNX_CPU_OPERATOR_KERNEL(
     KernelDefBuilder().TypeConstraint("T", DataTypeImpl::GetTensorType<float>()),
     Scale<float>);
 
+ONNX_CPU_OPERATOR_KERNEL(
+    Erf,
+    9,
+    KernelDefBuilder().TypeConstraint("T", DataTypeImpl::GetTensorType<float>()),
+    Erf<float>);
+
 template <typename T>
 Status Add<T>::Compute(OpKernelContext* context) const {
   return BroadcastTwo<T, T>(
@@ -874,4 +881,16 @@ Status Scale<float>::Compute(OpKernelContext* ctx) const {
   return Status::OK();
 }
 
+template <>
+Status Erf<float>::Compute(OpKernelContext* context) const {
+  auto X_ptr = context->Input<Tensor>(0);
+  ONNXRUNTIME_ENFORCE(X_ptr != nullptr);
+  auto& X = *X_ptr;
+  auto& Y = *context->Output(0, X.Shape());
+
+  EigenMap<float>(Y) = EigenMap<float>(X).array().erf();
+
+  return Status::OK();
+}
+
 }  // namespace onnxruntime
diff --git a/onnxruntime/core/providers/cpu/math/element_wise_ops.h b/onnxruntime/core/providers/cpu/math/element_wise_ops.h
index 912d352f39e24..feaa5c0a82b13 100644
--- a/onnxruntime/core/providers/cpu/math/element_wise_ops.h
+++ b/onnxruntime/core/providers/cpu/math/element_wise_ops.h
@@ -317,6 +317,15 @@ class Scale final : public OpKernel {
   float scale_;
 };
 
+template <typename T>
+class Erf final : public OpKernel {
+ public:
+  Erf(const OpKernelInfo& info) : OpKernel(info) {
+  }
+
+  Status Compute(OpKernelContext* context) const override;
+};
+
 template <typename T>
 auto MakeEigenArrayMap(Tensor& t) { return EigenVectorArrayMap<T>(t.template MutableData<T>(), t.Shape().Size()); }
 template <typename T>
diff --git a/onnxruntime/test/onnx/main.cc b/onnxruntime/test/onnx/main.cc
index 4871e1102ceed..77f717987e026 100644
--- a/onnxruntime/test/onnx/main.cc
+++ b/onnxruntime/test/onnx/main.cc
@@ -326,7 +326,6 @@ int real_main(int argc, char* argv[]) {
       {"acosh_example", "opset 9 not supported yet"},
       {"atanh_example", "opset 9 not supported yet"},
       {"sign_model", "opset 9 not supported yet"},
-      {"erf", "opset 9 not supported yet"},
       {"sign", "opset 9 not supported yet"},
       {"scatter_with_axis", "opset 9 not supported yet"},
       {"scatter_without_axis", "opset 9 not supported yet"},
diff --git a/onnxruntime/test/providers/cpu/math/element_wise_ops_test.cc b/onnxruntime/test/providers/cpu/math/element_wise_ops_test.cc
index e6eaf6c99e782..e8845542390fe 100644
--- a/onnxruntime/test/providers/cpu/math/element_wise_ops_test.cc
+++ b/onnxruntime/test/providers/cpu/math/element_wise_ops_test.cc
@@ -869,6 +869,13 @@ TEST(MathOpTest, Scale_Default) {
   test.Run();
 }
 
+TEST(MathOpTest, Erf) {
+  OpTester test("Erf", 9);
+  std::vector<int64_t> dims{2, 2};
+  test.AddInput<float>("A", dims, {0.5f, 1.0f, 0.7f, 2.0f});
+  test.AddOutput<float>("B", dims, {0.5204999f, 0.8427008f, 0.6778012f, 0.9953223f});
+  test.Run();
+}
 }  // namespace test
 
 }  // namespace onnxruntime