diff --git a/gallery/how_to/work_with_microtvm/install_cmsis.rst b/gallery/how_to/work_with_microtvm/install_cmsis.rst
index 2f1d2fb1189a5..13286b1b54f60 100644
--- a/gallery/how_to/work_with_microtvm/install_cmsis.rst
+++ b/gallery/how_to/work_with_microtvm/install_cmsis.rst
@@ -33,3 +33,7 @@ Install CMSIS-NN
         wget ${CMSIS_URL} -O "${DOWNLOAD_PATH}"
         tar -xf "${DOWNLOAD_PATH}" -C ${CMSIS_PATH} --strip-components=1
         rm ${DOWNLOAD_PATH}
+
+        CMSIS_NN_TAG="v4.0.0"
+        CMSIS_NN_URL="https://github.com/ARM-software/CMSIS-NN.git"
+        git clone ${CMSIS_NN_URL} --branch ${CMSIS_NN_TAG} --single-branch ${CMSIS_PATH}/CMSIS-NN
diff --git a/gallery/how_to/work_with_microtvm/micro_aot.py b/gallery/how_to/work_with_microtvm/micro_aot.py
index 81109b2965ef7..c1b29ba5c5820 100644
--- a/gallery/how_to/work_with_microtvm/micro_aot.py
+++ b/gallery/how_to/work_with_microtvm/micro_aot.py
@@ -41,7 +41,7 @@
 # By default, this tutorial runs on x86 CPU using TVM's C runtime. If you would like
 # to run on real Zephyr hardware, you must export the `TVM_MICRO_USE_HW` environment
 # variable. Otherwise (if you are using the C runtime), you can skip installing
-# Zephyr and CMSIS-NN. It takes ~20 minutes to install both of them.
+# Zephyr. It takes ~20 minutes to install Zephyr.
 use_physical_hw = bool(os.getenv("TVM_MICRO_USE_HW"))
 
 ######################################################################
@@ -49,11 +49,6 @@
 #     .. include:: ../../../../gallery/how_to/work_with_microtvm/install_zephyr.rst
 #
 
-######################################################################
-#
-#     .. include:: ../../../../gallery/how_to/work_with_microtvm/install_cmsis.rst
-#
-
 ######################################################################
 # Import Python dependencies
 # -------------------------------
@@ -159,7 +154,6 @@
         "board": BOARD,
         "serial_number": SERIAL,
         "config_main_stack_size": 4096,
-        "cmsis_path": os.getenv("CMSIS_PATH", default="/content/cmsis"),
         "zephyr_base": os.getenv("ZEPHYR_BASE", default="/content/zephyrproject/zephyr"),
     }
 
diff --git a/gallery/how_to/work_with_microtvm/micro_autotune.py b/gallery/how_to/work_with_microtvm/micro_autotune.py
index 9edb9ae75e7fb..9be257a57ac57 100644
--- a/gallery/how_to/work_with_microtvm/micro_autotune.py
+++ b/gallery/how_to/work_with_microtvm/micro_autotune.py
@@ -33,7 +33,7 @@
 #
 
 
-# You can skip the following two sections (installing Zephyr and CMSIS-NN) if the following flag is False.
+# You can skip the following section (installing Zephyr) if the following flag is False.
 # Installing Zephyr takes ~20 min.
 import os
 
@@ -44,10 +44,6 @@
 #     .. include:: ../../../../gallery/how_to/work_with_microtvm/install_zephyr.rst
 #
 
-######################################################################
-#
-#     .. include:: ../../../../gallery/how_to/work_with_microtvm/install_cmsis.rst
-#
 
 ######################################################################
 # Import Python dependencies
diff --git a/gallery/how_to/work_with_microtvm/micro_tflite.py b/gallery/how_to/work_with_microtvm/micro_tflite.py
index 86e5d6b4b1aec..0770d472c9b81 100644
--- a/gallery/how_to/work_with_microtvm/micro_tflite.py
+++ b/gallery/how_to/work_with_microtvm/micro_tflite.py
@@ -36,7 +36,7 @@
 # By default, this tutorial runs on x86 CPU using TVM's C runtime. If you would like
 # to run on real Zephyr hardware, you must export the `TVM_MICRO_USE_HW` environment
 # variable. Otherwise (if you are using the C runtime), you can skip installing
-# Zephyr and CMSIS-NN. It takes ~20 minutes to install both of them.
+# Zephyr. It takes ~20 minutes to install Zephyr.
 use_physical_hw = bool(os.getenv("TVM_MICRO_USE_HW"))
 
 ######################################################################
@@ -44,11 +44,6 @@
 #     .. include:: ../../../../gallery/how_to/work_with_microtvm/install_zephyr.rst
 #
 
-######################################################################
-#
-#     .. include:: ../../../../gallery/how_to/work_with_microtvm/install_cmsis.rst
-#
-
 ######################################################################
 # Import Python dependencies
 # -------------------------------
@@ -219,7 +214,6 @@
         "board": BOARD,
         "serial_number": SERIAL,
         "config_main_stack_size": 4096,
-        "cmsis_path": os.getenv("CMSIS_PATH", default="/content/cmsis"),
         "zephyr_base": os.getenv("ZEPHYR_BASE", default="/content/zephyrproject/zephyr"),
     }
 
diff --git a/src/target/parsers/aprofile.cc b/src/target/parsers/aprofile.cc
index 2fd5fe71e617f..6b0712461026c 100644
--- a/src/target/parsers/aprofile.cc
+++ b/src/target/parsers/aprofile.cc
@@ -134,15 +134,17 @@ static TargetFeatures GetFeatures(TargetJSON target) {
 }
 
 static Array<String> MergeKeys(Optional<Array<String>> existing_keys) {
-  const String kExtraKey = "arm_cpu";
+  const Array<String> kExtraKeys = {"arm_cpu", "cpu"};
 
   if (!existing_keys) {
-    return {kExtraKey};
+    return kExtraKeys;
   }
 
   Array<String> keys = existing_keys.value();
-  if (std::find(keys.begin(), keys.end(), kExtraKey) == keys.end()) {
-    keys.push_back(kExtraKey);
+  for (String key : kExtraKeys) {
+    if (std::find(keys.begin(), keys.end(), key) == keys.end()) {
+      keys.push_back(key);
+    }
   }
   return keys;
 }
diff --git a/tests/cpp/target/parsers/aprofile_test.cc b/tests/cpp/target/parsers/aprofile_test.cc
index 0382e7a84bd7e..ffbc5fd431a71 100644
--- a/tests/cpp/target/parsers/aprofile_test.cc
+++ b/tests/cpp/target/parsers/aprofile_test.cc
@@ -48,8 +48,9 @@ static TargetFeatures ParseTargetWithAttrs(String mcpu, String mtriple, Array<St
 TEST(AProfileParser, ParseTargetKeys) {
   TargetJSON target = ParseTarget({});
   Array<String> keys = Downcast<Array<String>>(target.at("keys"));
-  ASSERT_EQ(keys.size(), 1);
+  ASSERT_EQ(keys.size(), 2);
   ASSERT_EQ(keys[0], "arm_cpu");
+  ASSERT_EQ(keys[1], "cpu");
 }
 
 TEST(AProfileParser, ParseTargetWithExistingKeys) {