diff --git a/cmake/flags.cmake b/cmake/flags.cmake
index e26d8d9df386e65137aa83cc60a43bfeabf7a4a6..b27eb71550b68b5c27e47bf067ae0df329bbd628 100644
--- a/cmake/flags.cmake
+++ b/cmake/flags.cmake
@@ -9,10 +9,12 @@ function(CheckCompilerCXX11Flag)
         if(${CMAKE_CXX_COMPILER_VERSION} VERSION_LESS 4.8)
             message(FATAL_ERROR "Unsupported GCC version. GCC >= 4.8 required.")
         endif()
-        # TODO(qijun) gcc 4.9 or later versions raise SEGV due to the optimization problem.
-        # Use Debug mode instead for now.
-        if(CMAKE_CXX_COMPILER_VERSION VERSION_GREATER 4.9 OR CMAKE_CXX_COMPILER_VERSION VERSION_EQUAL 4.9) 
-            set(CMAKE_BUILD_TYPE "Debug" CACHE STRING "" FORCE)
+        if(NOT ANDROID)
+            # TODO(qijun) gcc 4.9 or later versions raise SEGV due to the optimization problem.
+            # Use Debug mode instead for now.
+            if(CMAKE_CXX_COMPILER_VERSION VERSION_GREATER 4.9 OR CMAKE_CXX_COMPILER_VERSION VERSION_EQUAL 4.9)
+                set(CMAKE_BUILD_TYPE "Debug" CACHE STRING "" FORCE)
+            endif()
         endif()
     elseif(CMAKE_CXX_COMPILER_ID STREQUAL "AppleClang" OR CMAKE_CXX_COMPILER_ID STREQUAL "Clang")
         # cmake >= 3.0 compiler id "AppleClang" on Mac OS X, otherwise "Clang"
diff --git a/paddle/capi/gradient_machine.cpp b/paddle/capi/gradient_machine.cpp
index e2d2d30ddcd205c26e64243fcf1c13642f19277e..f7ad30f3bf6e2ba83c099c0129068d1d77c0d628 100644
--- a/paddle/capi/gradient_machine.cpp
+++ b/paddle/capi/gradient_machine.cpp
@@ -54,6 +54,31 @@ paddle_error paddle_gradient_machine_create_for_inference(
   return kPD_NO_ERROR;
 }
 
+paddle_error paddle_gradient_machine_create_for_inference_with_parameters(
+    paddle_gradient_machine* machine, void* mergedModel, uint64_t size) {
+  if (mergedModel == nullptr) return kPD_NULLPTR;
+  std::istringstream is(std::string(static_cast<char*>(mergedModel), size));
+  int64_t modelConfigSize = 0;
+  is.read((char*)(&modelConfigSize), sizeof(modelConfigSize));
+  std::string modelConfigProtobuf;
+  modelConfigProtobuf.resize(modelConfigSize);
+  is.read(&modelConfigProtobuf[0], modelConfigSize);
+  paddle::TrainerConfig config;
+  if (!config.ParseFromString(modelConfigProtobuf) || !config.IsInitialized()) {
+    return kPD_PROTOBUF_ERROR;
+  }
+  auto ptr = new paddle::capi::CGradientMachine();
+  ptr->machine.reset(paddle::GradientMachine::create(
+      config.model_config(), CREATE_MODE_TESTING, {paddle::PARAMETER_VALUE}));
+  std::vector<paddle::ParameterPtr>& parameters = ptr->machine->getParameters();
+  for (auto& para : parameters) {
+    para->load(is);
+  }
+
+  *machine = ptr;
+  return kPD_NO_ERROR;
+}
+
 paddle_error paddle_gradient_machine_destroy(paddle_gradient_machine machine) {
   delete cast(machine);
   return kPD_NO_ERROR;
diff --git a/paddle/capi/gradient_machine.h b/paddle/capi/gradient_machine.h
index 242683905080326c44890cc411895c6951c89a0c..2205e0e23aaeb097021a882570f345a7fa0e5ffa 100644
--- a/paddle/capi/gradient_machine.h
+++ b/paddle/capi/gradient_machine.h
@@ -36,6 +36,18 @@ typedef void* paddle_gradient_machine;
 PD_API paddle_error paddle_gradient_machine_create_for_inference(
     paddle_gradient_machine* machine, void* modelConfigProtobuf, int size);
 
+/**
+ * @brief Create a gradient machine used for model inference, using config with
+ *        parameters which is generated by `paddle merge_model`.
+ * @param [out] machine that used for model inference.
+ * @param [in] mergedModel
+ * @param [in] size
+ * @return paddle_error
+ */
+PD_API paddle_error
+paddle_gradient_machine_create_for_inference_with_parameters(
+    paddle_gradient_machine* machine, void* mergedModel, uint64_t size);
+
 /**
  * @brief Load parameter from disk.
  * @param machine Gradient Machine.
diff --git a/paddle/gserver/gradientmachines/NeuralNetwork.cpp b/paddle/gserver/gradientmachines/NeuralNetwork.cpp
index 148296d20bdaeece2194727a2111d7fc6cb5ed55..cfa80a89365af5111746eec9599d16e37532a9f7 100644
--- a/paddle/gserver/gradientmachines/NeuralNetwork.cpp
+++ b/paddle/gserver/gradientmachines/NeuralNetwork.cpp
@@ -24,8 +24,6 @@ limitations under the License. */
 #include "paddle/gserver/layers/AgentLayer.h"
 #include "paddle/utils/Stat.h"
 
-#include <iostream>
-
 namespace paddle {
 void parameterInitNN(int paramId,
                      Parameter* para,