diff --git a/tensorflow/lite/tools/benchmark/BUILD b/tensorflow/lite/tools/benchmark/BUILD
index b0eaa4b5a06..8d41299fa57 100644
--- a/tensorflow/lite/tools/benchmark/BUILD
+++ b/tensorflow/lite/tools/benchmark/BUILD
@@ -124,6 +124,7 @@ cc_library(
         ":hexagon_delegate_provider",
         ":logging",
         ":nnapi_delegate_provider",
+        ":xnnpack_delegate_provider",
         "//tensorflow/lite:framework",
         "//tensorflow/lite:string_util",
         "//tensorflow/lite/experimental/ruy/profiler",
@@ -260,6 +261,20 @@ cc_library(
     alwayslink = 1,
 )
 
+cc_library(
+    name = "xnnpack_delegate_provider",
+    srcs = ["xnnpack_delegate_provider.cc"],
+    copts = tflite_copts(),
+    visibility = ["//visibility:public"],
+    deps = [
+        ":benchmark_model_lib",
+        ":delegate_provider_hdr",
+        ":logging",
+        "//tensorflow/lite/delegates/xnnpack:xnnpack_delegate",
+    ],
+    alwayslink = 1,
+)
+
 cc_library(
     name = "benchmark_utils",
     srcs = [
diff --git a/tensorflow/lite/tools/benchmark/README.md b/tensorflow/lite/tools/benchmark/README.md
index b8011cfbfe3..f510a3d06b6 100644
--- a/tensorflow/lite/tools/benchmark/README.md
+++ b/tensorflow/lite/tools/benchmark/README.md
@@ -34,6 +34,8 @@ and the following optional parameters:
 *   `run_delay`: `float` (default=-1.0) \
     The delay in seconds between subsequent benchmark runs. Non-positive values
     mean use no delay.
+*   `use_xnnpack`: `bool` (default=false) \
+    Whether to use the XNNPack delegate.
 *   `use_hexagon`: `bool` (default=false) \
     Whether to use the Hexagon delegate. Not all devices may support the Hexagon
     delegate, refer to the TensorFlow Lite documentation for more information
diff --git a/tensorflow/lite/tools/benchmark/benchmark_performance_options.cc b/tensorflow/lite/tools/benchmark/benchmark_performance_options.cc
index e286b7c9b0c..eccd389af6f 100644
--- a/tensorflow/lite/tools/benchmark/benchmark_performance_options.cc
+++ b/tensorflow/lite/tools/benchmark/benchmark_performance_options.cc
@@ -79,6 +79,12 @@ void MultiRunStatsRecorder::OnBenchmarkStart(const BenchmarkParams& params) {
   // requires C++11.
   std::stringstream sstm;
   sstm << "cpu w/ " << params.Get<int32_t>("num_threads") << " threads";
+
+  // Handle cases run on CPU w/ the xnnpack delegate
+  if (params.Get<bool>("use_xnnpack")) {
+    sstm << " (xnnpack)";
+  }
+
   current_run_name_ = sstm.str();
 }
 
@@ -238,6 +244,7 @@ void BenchmarkPerformanceOptions::ResetPerformanceOptions() {
 #if defined(TFLITE_ENABLE_HEXAGON)
   single_option_run_params_->Set<bool>("use_hexagon", false);
 #endif
+  single_option_run_params_->Set<bool>("use_xnnpack", false);
 }
 
 void BenchmarkPerformanceOptions::CreatePerformanceOptions() {
@@ -260,6 +267,13 @@ void BenchmarkPerformanceOptions::CreatePerformanceOptions() {
       BenchmarkParams params;
       params.AddParam("num_threads", BenchmarkParam::Create<int32_t>(count));
       all_run_params_.emplace_back(std::move(params));
+
+      BenchmarkParams xnnpack_params;
+      xnnpack_params.AddParam("use_xnnpack",
+                              BenchmarkParam::Create<bool>(true));
+      xnnpack_params.AddParam("num_threads",
+                              BenchmarkParam::Create<int32_t>(count));
+      all_run_params_.emplace_back(std::move(xnnpack_params));
     }
   }
 
diff --git a/tensorflow/lite/tools/benchmark/benchmark_test.cc b/tensorflow/lite/tools/benchmark/benchmark_test.cc
index 463b3b4117b..19e0b71893e 100644
--- a/tensorflow/lite/tools/benchmark/benchmark_test.cc
+++ b/tensorflow/lite/tools/benchmark/benchmark_test.cc
@@ -62,6 +62,7 @@ BenchmarkParams CreateParams(int32_t num_runs, float min_secs, float max_secs,
   params.AddParam("input_layer_value_range",
                   BenchmarkParam::Create<std::string>(""));
   params.AddParam("use_hexagon", BenchmarkParam::Create<bool>(false));
+  params.AddParam("use_xnnpack", BenchmarkParam::Create<bool>(false));
   params.AddParam("use_nnapi", BenchmarkParam::Create<bool>(false));
   params.AddParam("allow_fp16", BenchmarkParam::Create<bool>(false));
   params.AddParam("require_full_delegation",
diff --git a/tensorflow/lite/tools/benchmark/xnnpack_delegate_provider.cc b/tensorflow/lite/tools/benchmark/xnnpack_delegate_provider.cc
new file mode 100644
index 00000000000..63270fc2cd4
--- /dev/null
+++ b/tensorflow/lite/tools/benchmark/xnnpack_delegate_provider.cc
@@ -0,0 +1,72 @@
+/* Copyright 2020 The TensorFlow Authors. All Rights Reserved.
+
+Licensed under the Apache License, Version 2.0 (the "License");
+you may not use this file except in compliance with the License.
+You may obtain a copy of the License at
+
+    http://www.apache.org/licenses/LICENSE-2.0
+
+Unless required by applicable law or agreed to in writing, software
+distributed under the License is distributed on an "AS IS" BASIS,
+WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+See the License for the specific language governing permissions and
+limitations under the License.
+==============================================================================*/
+#include <string>
+
+#include "tensorflow/lite/delegates/xnnpack/xnnpack_delegate.h"
+#include "tensorflow/lite/tools/benchmark/benchmark_model.h"
+#include "tensorflow/lite/tools/benchmark/delegate_provider.h"
+#include "tensorflow/lite/tools/benchmark/logging.h"
+
+namespace tflite {
+namespace benchmark {
+
+class XnnpackDelegateProvider : public DelegateProvider {
+ public:
+  std::vector<Flag> CreateFlags(BenchmarkParams* params) const final;
+
+  void AddParams(BenchmarkParams* params) const final;
+
+  void LogParams(const BenchmarkParams& params) const final;
+
+  TfLiteDelegatePtr CreateTfLiteDelegate(
+      const BenchmarkParams& params) const final;
+
+  std::string GetName() const final { return "XNNPACK"; }
+};
+REGISTER_DELEGATE_PROVIDER(XnnpackDelegateProvider);
+
+std::vector<Flag> XnnpackDelegateProvider::CreateFlags(
+    BenchmarkParams* params) const {
+  std::vector<Flag> flags = {
+      CreateFlag<bool>("use_xnnpack", params, "use XNNPack")};
+  return flags;
+}
+
+void XnnpackDelegateProvider::AddParams(BenchmarkParams* params) const {
+  params->AddParam("use_xnnpack", BenchmarkParam::Create<bool>(false));
+}
+
+void XnnpackDelegateProvider::LogParams(const BenchmarkParams& params) const {
+  TFLITE_LOG(INFO) << "Use xnnpack : [" << params.Get<bool>("use_xnnpack")
+                   << "]";
+}
+
+TfLiteDelegatePtr XnnpackDelegateProvider::CreateTfLiteDelegate(
+    const BenchmarkParams& params) const {
+  TfLiteDelegatePtr delegate(nullptr, [](TfLiteDelegate*) {});
+  if (params.Get<bool>("use_xnnpack")) {
+    TfLiteXNNPackDelegateOptions options =
+        TfLiteXNNPackDelegateOptionsDefault();
+    const auto num_threads = params.Get<int32_t>("num_threads");
+    // Note that we don't want to use the thread pool for num_threads == 1.
+    options.num_threads = num_threads > 1 ? num_threads : 0;
+    delegate = TfLiteDelegatePtr(TfLiteXNNPackDelegateCreate(&options),
+                                 &TfLiteXNNPackDelegateDelete);
+  }
+  return delegate;
+}
+
+}  // namespace benchmark
+}  // namespace tflite