diff --git a/src/plugins/intel_gpu/include/intel_gpu/graph/serialization/binary_buffer.hpp b/src/plugins/intel_gpu/include/intel_gpu/graph/serialization/binary_buffer.hpp
index 7880b79a85eb4d..182865306e4611 100644
--- a/src/plugins/intel_gpu/include/intel_gpu/graph/serialization/binary_buffer.hpp
+++ b/src/plugins/intel_gpu/include/intel_gpu/graph/serialization/binary_buffer.hpp
@@ -14,7 +14,8 @@
 namespace cldnn {
 class BinaryOutputBuffer : public OutputBuffer<BinaryOutputBuffer> {
 public:
-    BinaryOutputBuffer(std::ostream& stream) : OutputBuffer<BinaryOutputBuffer>(this), stream(stream) {}
+    BinaryOutputBuffer(std::ostream& stream)
+    : OutputBuffer<BinaryOutputBuffer>(this), stream(stream), _impl_params(nullptr) {}
 
     void write(void const * data, std::streamsize size) {
         auto const written_size = stream.rdbuf()->sputn(reinterpret_cast<const char*>(data), size);
@@ -32,7 +33,8 @@ class BinaryOutputBuffer : public OutputBuffer<BinaryOutputBuffer> {
 
 class BinaryInputBuffer : public InputBuffer<BinaryInputBuffer> {
 public:
-    BinaryInputBuffer(std::istream& stream, engine& engine) : InputBuffer(this, engine), stream(stream) {}
+    BinaryInputBuffer(std::istream& stream, engine& engine)
+    : InputBuffer(this, engine), stream(stream), _impl_params(nullptr), _network(nullptr) {}
 
     void read(void* const data, std::streamsize size) {
         auto const read_size = stream.rdbuf()->sgetn(reinterpret_cast<char*>(data), size);
@@ -42,6 +44,8 @@ class BinaryInputBuffer : public InputBuffer<BinaryInputBuffer> {
 
     void setKernlImplParams(void* impl_params) { _impl_params = impl_params; }
     void* getKernlImplParams() const { return _impl_params; }
+    void setNetwork(void* network) { _network = network; }
+    void* getNetwork() const { return _network; }
 
     std::streampos tellg() { return stream.tellg(); }
     void seekg(std::streampos pos) { stream.seekg(pos); }
@@ -49,6 +53,7 @@ class BinaryInputBuffer : public InputBuffer<BinaryInputBuffer> {
 private:
     std::istream& stream;
     void* _impl_params;
+    void* _network;
 };
 
 template <typename T>
diff --git a/src/plugins/intel_gpu/src/graph/data.cpp b/src/plugins/intel_gpu/src/graph/data.cpp
index 10be2a3504e81a..16e0edb6d2d033 100644
--- a/src/plugins/intel_gpu/src/graph/data.cpp
+++ b/src/plugins/intel_gpu/src/graph/data.cpp
@@ -85,15 +85,24 @@ void data_inst::load(BinaryInputBuffer& ib) {
 
     size_t data_size;
     ib >> make_data(&data_size, sizeof(size_t));
-    _outputs[0] = get_network().get_memory_pool().get_memory(output_layout, _allocation_type, false);
 
-    if (_allocation_type == allocation_type::usm_host || _allocation_type == allocation_type::usm_shared) {
-        ib >> make_data(_outputs[0]->buffer_ptr(), data_size);
+    if (ib.getNetwork()) {
+        const network* primary_network = reinterpret_cast<network*>(ib.getNetwork());
+        _outputs[0] = primary_network->get_primitive(id())->output_memory_ptr();
+        auto pos = ib.tellg();
+        pos += data_size;
+        ib.seekg(pos);
     } else {
-        std::vector<uint8_t> _buf;
-        _buf.resize(data_size);
-        ib >> make_data(_buf.data(), data_size);
-        _outputs[0]->copy_from(get_network().get_stream(), _buf.data());
+        _outputs[0] = get_network().get_memory_pool().get_memory(output_layout, _allocation_type, false);
+
+        if (_allocation_type == allocation_type::usm_host || _allocation_type == allocation_type::usm_shared) {
+            ib >> make_data(_outputs[0]->buffer_ptr(), data_size);
+        } else {
+            std::vector<uint8_t> _buf;
+            _buf.resize(data_size);
+            ib >> make_data(_buf.data(), data_size);
+            _outputs[0]->copy_from(get_network().get_stream(), _buf.data());
+        }
     }
 }
 
diff --git a/src/plugins/intel_gpu/src/graph/primitive_inst.cpp b/src/plugins/intel_gpu/src/graph/primitive_inst.cpp
index 697384adb3b0f2..2eb6b3ec55a776 100644
--- a/src/plugins/intel_gpu/src/graph/primitive_inst.cpp
+++ b/src/plugins/intel_gpu/src/graph/primitive_inst.cpp
@@ -518,7 +518,7 @@ void primitive_inst::rebuild_exec_deps(
 primitive_inst::primitive_inst(network& network)
     : _network(network)
     , _node(nullptr)
-    , _impl_params(nullptr)
+    , _impl_params(make_unique<kernel_impl_params>())
     , _impl(nullptr)
     , _dynamic_impl(nullptr)
     , _outputs({memory::ptr()})
@@ -1149,8 +1149,6 @@ int32_t primitive_inst::get_index_in_deps(memory::cptr arg) const {
 }
 
 void primitive_inst::load(cldnn::BinaryInputBuffer& ib) {
-    _impl_params.release();
-    _impl_params = make_unique<kernel_impl_params>();
     _impl_params->load(ib);
     ib.setKernlImplParams(_impl_params.get());
 
diff --git a/src/plugins/intel_gpu/src/plugin/compiled_model.cpp b/src/plugins/intel_gpu/src/plugin/compiled_model.cpp
index d638e38dec1304..c699b379984c94 100644
--- a/src/plugins/intel_gpu/src/plugin/compiled_model.cpp
+++ b/src/plugins/intel_gpu/src/plugin/compiled_model.cpp
@@ -241,6 +241,9 @@ CompiledModel::CompiledModel(std::istream& networkModel, InferenceEngine::Remote
         ib.seekg(pos);
         auto graph = std::make_shared<Graph>(ib, context_impl, m_config, n);
         m_graphs.push_back(graph);
+        if (n == 0) {
+            ib.setNetwork(graph->GetNetwork().get());
+        }
     }
 }