diff --git a/inference-engine/src/gna_plugin/backend/am_intel_dnn.cpp b/inference-engine/src/gna_plugin/backend/am_intel_dnn.cpp
index 43b0fa34642..b41166c7715 100644
--- a/inference-engine/src/gna_plugin/backend/am_intel_dnn.cpp
+++ b/inference-engine/src/gna_plugin/backend/am_intel_dnn.cpp
@@ -1453,12 +1453,6 @@ void GNAPluginNS::backend::AMIntelDNN::InitGNAStruct(intel_nnet_type_t *ptr_nnet
                                 comp.op.conv1D.num_feature_maps * comp.op.conv1D.num_feature_map_columns),
                         nullptr);
 
-                // TODO: GNA2: We have to explicitly enforce to use Legacy CNN
-                snprintf(
-                        const_cast<char*>(gnaOperation->Operands[1]->Layout),
-                        sizeof(gnaOperation->Operands[1]->Layout) / sizeof(char),
-                        "GNA1");
-
                 AdvanceCnnOperationIfAllApplied(component, i, gnaOperation);
 #else
                 pLayer->nInputRows = component[i].num_rows_in;
diff --git a/inference-engine/src/gna_plugin/gna_device.cpp b/inference-engine/src/gna_plugin/gna_device.cpp
index ce15f2b895b..7cd9fc27127 100644
--- a/inference-engine/src/gna_plugin/gna_device.cpp
+++ b/inference-engine/src/gna_plugin/gna_device.cpp
@@ -90,8 +90,22 @@ uint32_t GNADeviceHelper::propagate(const uint32_t requestConfigId, Gna2Accelera
     return reqId;
 }
 
-uint32_t GNADeviceHelper::createModel(const Gna2Model& gnaModel) const {
+void GNADeviceHelper::enforceLegacyCnns(Gna2Model& gnaModel) {
+    for (uint32_t i = 0; i < gnaModel.NumberOfOperations; i++) {
+        if (gnaModel.Operations->Type == Gna2OperationTypeConvolution) {
+            snprintf(
+                const_cast<char*>(gnaModel.Operations[i].Operands[1]->Layout),
+                sizeof(gnaModel.Operations[i].Operands[1]->Layout) / sizeof(char),
+                "GNA1");
+        }
+    }
+}
+
+uint32_t GNADeviceHelper::createModel(Gna2Model& gnaModel) const {
     uint32_t modelId;
+    if (isUpTo20GnaDevice()) {
+        enforceLegacyCnns(gnaModel);
+    }
     const auto status = Gna2ModelCreate(nGnaDeviceIndex, &gnaModel, &modelId);
 
     checkGna2Status(status, gnaModel);
@@ -108,7 +122,8 @@ uint32_t GNADeviceHelper::createRequestConfig(const uint32_t model_id) {
     auto status = Gna2RequestConfigCreate(model_id, &reqConfId);
     checkGna2Status(status);
     if (gna2HwConsistency != Gna2DeviceVersionSoftwareEmulation) {
-        status = Gna2RequestConfigEnableHardwareConsistency(reqConfId, gna2HwConsistency);
+        status = Gna2RequestConfigEnableHardwareConsistency(reqConfId,
+            isUpTo20GnaDevice() ? gna2HwConsistency : detectedGnaDevVersion);
         checkGna2Status(status);
     }
     status = Gna2InstrumentationConfigAssignToRequestConfig(instrumentationConfigId, reqConfId);
diff --git a/inference-engine/src/gna_plugin/gna_device.hpp b/inference-engine/src/gna_plugin/gna_device.hpp
index 0f71772d62e..d067ce9f410 100644
--- a/inference-engine/src/gna_plugin/gna_device.hpp
+++ b/inference-engine/src/gna_plugin/gna_device.hpp
@@ -110,7 +110,7 @@ public:
     void propagateSync(const uint32_t requestConfigId, Gna2AccelerationMode gna2AccelerationMode);
     uint32_t propagate(const uint32_t requestConfigId, Gna2AccelerationMode gna2AccelerationMode);
 #if GNA_LIB_VER == 2
-    uint32_t createModel(const Gna2Model& gnaModel) const;
+    uint32_t createModel(Gna2Model& gnaModel) const;
 #else
     uint32_t createModel(const intel_nnet_type_t& intel_nnet_type);
 #endif
@@ -119,6 +119,9 @@ public:
     bool hasGnaHw() const {
         return Gna2DeviceVersionSoftwareEmulation != detectedGnaDevVersion;
     }
+    bool isUpTo20GnaDevice() const {
+        return detectedGnaDevVersion <= Gna2DeviceVersion2_0;
+    }
     static void checkGna2Status(Gna2Status status);
     static void checkGna2Status(Gna2Status status, const Gna2Model& gnaModel);
 #endif
@@ -166,6 +169,8 @@ public:
     static const std::map <Gna2ErrorType, const std::string> errorReasons;
     static const std::map <Gna2OperationType, const std::string> operationTypes;
     static const std::map <const std::pair<Gna2OperationType, int32_t>, const std::string > operandTypes;
+
+    static void enforceLegacyCnns(Gna2Model& gnaModel);
 #endif
     void setOMPThreads(uint8_t const n_threads);
 
diff --git a/inference-engine/src/gna_plugin/gna_plugin.cpp b/inference-engine/src/gna_plugin/gna_plugin.cpp
index 3905da7c039..d7b704f5c88 100644
--- a/inference-engine/src/gna_plugin/gna_plugin.cpp
+++ b/inference-engine/src/gna_plugin/gna_plugin.cpp
@@ -756,7 +756,7 @@ void GNAPlugin::createRequestConfigsForGnaModels() {
         return;
     }
     for (auto& model : gnaModels) {
-        const auto& gnaNnet = std::get<0>(model).get()->obj;
+        auto& gnaNnet = std::get<0>(model).get()->obj;
         const auto modelId = gnadevice->createModel(gnaNnet);
         const auto requestConfigId = gnadevice->createRequestConfig(modelId);
         gnaRequestConfigToRequestIdMap.push_back(std::make_tuple(requestConfigId, -1, InferenceEngine::BlobMap()));