diff --git a/frameworks/native/neural_network_runtime/device.h b/frameworks/native/neural_network_runtime/device.h
index 274b9796e26d928404db56702ce2b500218490f4..015e9e5d0380eff8e1a4515c12eb538b6ab9039b 100644
--- a/frameworks/native/neural_network_runtime/device.h
+++ b/frameworks/native/neural_network_runtime/device.h
@@ -49,6 +49,7 @@ public:
     virtual OH_NN_ReturnCode IsModelCacheSupported(bool& isSupported) = 0;
 
     virtual OH_NN_ReturnCode PrepareModel(std::shared_ptr<const mindspore::lite::LiteGraph> model,
+                                          const Buffer& quantBuffer,
                                           const ModelConfig& config,
                                           std::shared_ptr<PreparedModel>& preparedModel) = 0;
     virtual OH_NN_ReturnCode PrepareModel(const void* metaGraph,
@@ -72,4 +73,4 @@ public:
 };
 } // namespace NeuralNetworkRuntime
 } // namespace OHOS
-#endif // NEURAL_NETWORK_RUNTIME_DEVICE_H
\ No newline at end of file
+#endif // NEURAL_NETWORK_RUNTIME_DEVICE_H
diff --git a/frameworks/native/neural_network_runtime/hdi_device_v1_0.cpp b/frameworks/native/neural_network_runtime/hdi_device_v1_0.cpp
index ef69c9a13e9ee921de17b9627cfa5d857c3e72d6..87cd2eb9ef3be9efae984f80a895707ed8358971 100644
--- a/frameworks/native/neural_network_runtime/hdi_device_v1_0.cpp
+++ b/frameworks/native/neural_network_runtime/hdi_device_v1_0.cpp
@@ -238,7 +238,7 @@ OH_NN_ReturnCode HDIDeviceV1_0::IsModelCacheSupported(bool& isSupported)
 }
 
 OH_NN_ReturnCode HDIDeviceV1_0::PrepareModel(std::shared_ptr<const mindspore::lite::LiteGraph> model,
-    const ModelConfig& config, std::shared_ptr<PreparedModel>& preparedModel)
+    const Buffer& quantBuffer, const ModelConfig& config, std::shared_ptr<PreparedModel>& preparedModel)
 {
     if (model == nullptr) {
         LOGE("Model is nullptr, cannot prepare model.");
diff --git a/frameworks/native/neural_network_runtime/hdi_device_v1_0.h b/frameworks/native/neural_network_runtime/hdi_device_v1_0.h
index a7167ba177f7e1d9a7efb962557ae92f31962a19..fb0f2aeb41f0782ef1b0d7f36720c3642a53e16c 100644
--- a/frameworks/native/neural_network_runtime/hdi_device_v1_0.h
+++ b/frameworks/native/neural_network_runtime/hdi_device_v1_0.h
@@ -45,6 +45,7 @@ public:
     OH_NN_ReturnCode IsModelCacheSupported(bool& isSupported) override;
 
     OH_NN_ReturnCode PrepareModel(std::shared_ptr<const mindspore::lite::LiteGraph> model,
+                                  const Buffer& quantBuffer,
                                   const ModelConfig& config,
                                   std::shared_ptr<PreparedModel>& preparedModel) override;
     OH_NN_ReturnCode PrepareModel(const void* metaGraph,
@@ -76,4 +77,4 @@ private:
 };
 } // namespace NeuralNetworkRuntime
 } // namespace OHOS
-#endif // NEURAL_NETWORK_RUNTIME_HDI_DEVICE_V1_0_H
\ No newline at end of file
+#endif // NEURAL_NETWORK_RUNTIME_HDI_DEVICE_V1_0_H
diff --git a/frameworks/native/neural_network_runtime/hdi_device_v2_0.cpp b/frameworks/native/neural_network_runtime/hdi_device_v2_0.cpp
index 221bb6b0369f7c2b0af23cc96e6e9503a6a87918..09c92d0ef203c03faab799b7d8caca5603d83158 100644
--- a/frameworks/native/neural_network_runtime/hdi_device_v2_0.cpp
+++ b/frameworks/native/neural_network_runtime/hdi_device_v2_0.cpp
@@ -277,7 +277,7 @@ OH_NN_ReturnCode HDIDeviceV2_0::IsModelCacheSupported(bool& isSupported)
 }
 
 OH_NN_ReturnCode HDIDeviceV2_0::PrepareModel(std::shared_ptr<const mindspore::lite::LiteGraph> model,
-    const ModelConfig& config, std::shared_ptr<PreparedModel>& preparedModel)
+    const Buffer& quantBuffer, const ModelConfig& config, std::shared_ptr<PreparedModel>& preparedModel)
 {
     if (model == nullptr) {
         LOGE("Model is nullptr, cannot prepare model.");
diff --git a/frameworks/native/neural_network_runtime/hdi_device_v2_0.h b/frameworks/native/neural_network_runtime/hdi_device_v2_0.h
index b8dc5f8f0c58a9bd3204e6e0ee009c2ba93bb026..2c3aab5e0b7f940feb8fe8e2c82d3879e94687f2 100644
--- a/frameworks/native/neural_network_runtime/hdi_device_v2_0.h
+++ b/frameworks/native/neural_network_runtime/hdi_device_v2_0.h
@@ -45,6 +45,7 @@ public:
     OH_NN_ReturnCode IsModelCacheSupported(bool& isSupported) override;
 
     OH_NN_ReturnCode PrepareModel(std::shared_ptr<const mindspore::lite::LiteGraph> model,
+                                  const Buffer& quantBuffer,
                                   const ModelConfig& config,
                                   std::shared_ptr<PreparedModel>& preparedModel) override;
     OH_NN_ReturnCode PrepareModel(const void* metaGraph,
diff --git a/frameworks/native/neural_network_runtime/hdi_device_v2_1.cpp b/frameworks/native/neural_network_runtime/hdi_device_v2_1.cpp
index 72b5cc4558ca4899827987ae22467a5e13a0ede4..9b9f0f67253f500fbaeaabcba3509f4d23d5691b 100644
--- a/frameworks/native/neural_network_runtime/hdi_device_v2_1.cpp
+++ b/frameworks/native/neural_network_runtime/hdi_device_v2_1.cpp
@@ -277,7 +277,7 @@ OH_NN_ReturnCode HDIDeviceV2_1::IsModelCacheSupported(bool& isSupported)
 }
 
 OH_NN_ReturnCode HDIDeviceV2_1::PrepareModel(std::shared_ptr<const mindspore::lite::LiteGraph> model,
-    const ModelConfig& config, std::shared_ptr<PreparedModel>& preparedModel)
+    const Buffer& quantBuffer, const ModelConfig& config, std::shared_ptr<PreparedModel>& preparedModel)
 {
     if (model == nullptr) {
         LOGE("Model is nullptr, cannot prepare model.");
diff --git a/frameworks/native/neural_network_runtime/hdi_device_v2_1.h b/frameworks/native/neural_network_runtime/hdi_device_v2_1.h
index a2c4910f3bf8b46b9d52b8c4c879f5affaf52e78..06016b9e334d89e62d79a3c0e96c177b90c6ee31 100644
--- a/frameworks/native/neural_network_runtime/hdi_device_v2_1.h
+++ b/frameworks/native/neural_network_runtime/hdi_device_v2_1.h
@@ -45,6 +45,7 @@ public:
     OH_NN_ReturnCode IsModelCacheSupported(bool& isSupported) override;
 
     OH_NN_ReturnCode PrepareModel(std::shared_ptr<const mindspore::lite::LiteGraph> model,
+                                  const Buffer& quantBuffer,
                                   const ModelConfig& config,
                                   std::shared_ptr<PreparedModel>& preparedModel) override;
     OH_NN_ReturnCode PrepareModel(const void* metaGraph,
diff --git a/frameworks/native/neural_network_runtime/nnbackend.h b/frameworks/native/neural_network_runtime/nnbackend.h
index 826817010216f3bb3ddaafb8c373ee1d7f4d62c5..83a4af85fab211e0188faf4d8ae34b1cc7bea561 100644
--- a/frameworks/native/neural_network_runtime/nnbackend.h
+++ b/frameworks/native/neural_network_runtime/nnbackend.h
@@ -54,11 +54,11 @@ public:
     OH_NN_ReturnCode GetSupportedOperation(std::shared_ptr<const mindspore::lite::LiteGraph> model,
                                            std::vector<bool>& ops);
 
-private:
+protected:
     std::shared_ptr<Device> m_device;
     size_t m_backendID;
 };
 } // NeuralNetworkRuntime
 } // OHOS
 
-#endif // NEURAL_NETWORK_RUNTIME_NNBACKEND_H
\ No newline at end of file
+#endif // NEURAL_NETWORK_RUNTIME_NNBACKEND_H
diff --git a/frameworks/native/neural_network_runtime/nncompiler.cpp b/frameworks/native/neural_network_runtime/nncompiler.cpp
index bc7a40987b5e539b60ee913832b1da848bdcae84..56ab88632363f1d65b9d2a3514243a5a4cd9f8a4 100644
--- a/frameworks/native/neural_network_runtime/nncompiler.cpp
+++ b/frameworks/native/neural_network_runtime/nncompiler.cpp
@@ -377,7 +377,7 @@ OH_NN_ReturnCode NNCompiler::NormalBuild()
     ModelConfig config {m_enableFp16, static_cast<OH_NN_PerformanceMode>(m_performance),
         static_cast<OH_NN_Priority>(m_priority), m_isProfiling, m_cachePath, m_opLayouts};
     if (m_liteGraph != nullptr) {
-        ret = m_device->PrepareModel(m_liteGraph, config, m_preparedModel);
+        ret = m_device->PrepareModel(m_liteGraph, m_quantBuffer, config, m_preparedModel);
     }
     if (m_metaGraph != nullptr) {
         ret = m_device->PrepareModel(m_metaGraph, m_quantBuffer, config, m_preparedModel);
@@ -856,4 +856,4 @@ OH_NN_ReturnCode NNCompiler::DeserializedTensorsFromBuffer(
 }
 
 } // NeuralNetworkRuntime
-} // OHOS
\ No newline at end of file
+} // OHOS