diff --git a/cmake/external/lite.cmake b/cmake/external/lite.cmake
index 09be60ab4784408d0c2b9b73e6f2477089e9dcac..9e26cd4fa2dfe17e407ee05892d535c8c3d6d7ef 100644
--- a/cmake/external/lite.cmake
+++ b/cmake/external/lite.cmake
@@ -36,7 +36,7 @@ if (NOT LITE_SOURCE_DIR OR NOT LITE_BINARY_DIR)
   set(LITE_INSTALL_DIR ${THIRD_PARTY_PATH}/install/lite)
 
   if(NOT LITE_GIT_TAG)
-    set(LITE_GIT_TAG 68e64e0eb74cdd13383ae78caf889973499ebd14)
+    set(LITE_GIT_TAG d3a3a6931b6d22d504d21ba32b3ae972770e9204)
   endif()
 
   if(NOT CUDA_ARCH_NAME)
diff --git a/paddle/fluid/inference/analysis/ir_passes/lite_subgraph_pass.cc b/paddle/fluid/inference/analysis/ir_passes/lite_subgraph_pass.cc
index 2c454893a620360d33a790dc78bd4f33995c4632..4402d5c595a2370d7d6c451c8f051d18d239d207 100644
--- a/paddle/fluid/inference/analysis/ir_passes/lite_subgraph_pass.cc
+++ b/paddle/fluid/inference/analysis/ir_passes/lite_subgraph_pass.cc
@@ -272,6 +272,8 @@ void LiteSubgraphPass::SetUpEngine(
       paddle::lite_api::Place({target_type, PRECISION(kInt64)}),
       paddle::lite_api::Place({target_type, PRECISION(kFloat)}),
       paddle::lite_api::Place({TARGET(kHost), PRECISION(kFloat)}),
+      paddle::lite_api::Place({TARGET(kX86), precision_type}),
+      paddle::lite_api::Place({TARGET(kX86), PRECISION(kFloat)}),
   };
   config.cpu_math_library_num_threads = cpu_math_library_num_threads;
   config.xpu_l3_workspace_size = xpu_l3_workspace_size;
diff --git a/paddle/fluid/inference/lite/tensor_utils.cc b/paddle/fluid/inference/lite/tensor_utils.cc
index f108e7d3651ca42fc673523e8a9d5f7d2032ded2..bad4b7e50a71df8638f81be9f286ff051486bcbb 100644
--- a/paddle/fluid/inference/lite/tensor_utils.cc
+++ b/paddle/fluid/inference/lite/tensor_utils.cc
@@ -185,10 +185,8 @@ void InitDstTensor(paddle::lite_api::Tensor* dst,
 
 void InitDstTensor(framework::LoDTensor* dst,
                    const paddle::lite_api::Tensor& src) {
-  constexpr framework::proto::VarType::Type dtype =
-      framework::proto::VarType_Type_FP32;
   dst->mutable_data(inference::lite::utils::GetNativePlace(src.target()),
-                    dtype);
+                    GetNativePrecisionType(src.precision()));
   SetLoD(dst->mutable_lod(), src.lod());
 }
 
@@ -244,17 +242,17 @@ void TensorDataShare(paddle::lite_api::Tensor* dst, framework::LoDTensor* src) {
 
 template <>
 void TensorDataShare(framework::LoDTensor* dst, paddle::lite_api::Tensor* src) {
-  constexpr framework::proto::VarType::Type dtype =
-      framework::proto::VarType_Type_FP32;
   void* src_raw_data =
-      GetLiteTensorDataPtr(src, GetLitePrecisionType(dtype), src->target());
-  size_t memory_size = GetLiteTensorNumel(*src) * sizeof(float);
+      GetLiteTensorDataPtr(src, src->precision(), src->target());
+  size_t memory_size =
+      GetLiteTensorNumel(*src) *
+      framework::SizeOfType(GetNativePrecisionType(src->precision()));
   std::shared_ptr<memory::allocation::Allocation> holder(
       new memory::allocation::Allocation(src_raw_data, memory_size,
                                          GetNativePlace(src->target())));
   dst->Resize(paddle::framework::make_ddim(src->shape()));
   SetLoD(dst->mutable_lod(), src->lod());
-  dst->ResetHolderWithType(holder, dtype);
+  dst->ResetHolderWithType(holder, GetNativePrecisionType(src->precision()));
 }
 
 }  // namespace utils