diff --git a/src/inference/src/dev/core_impl.cpp b/src/inference/src/dev/core_impl.cpp
index c0a99d0fbb7..462a0b9f7cc 100644
--- a/src/inference/src/dev/core_impl.cpp
+++ b/src/inference/src/dev/core_impl.cpp
@@ -743,7 +743,7 @@ ov::SoPtr<ov::ICompiledModel> ov::CoreImpl::compile_model(const std::shared_ptr<
     // if auto-batching is applicable, the below function will patch the device name and config accordingly:
     auto model = apply_auto_batching(model_, deviceName, config_with_batch);
 
-    auto parsed = parseDeviceNameIntoConfig(deviceName, config_with_batch);
+    auto parsed = parseDeviceNameIntoConfig(deviceName, config_with_batch, is_proxy_device(device_name));
     auto plugin = get_plugin(parsed._deviceName);
     ov::SoPtr<ov::ICompiledModel> res;
     auto cacheManager = coreConfig.get_cache_config_for_device(plugin, parsed._config)._cacheManager;
@@ -776,7 +776,7 @@ ov::SoPtr<ov::ICompiledModel> ov::CoreImpl::compile_model(const std::shared_ptr<
     // if auto-batching is applicable, the below function will patch the device name and config accordingly:
     auto model = apply_auto_batching(model_, deviceName, config_with_batch);
 
-    auto parsed = parseDeviceNameIntoConfig(deviceName, config_with_batch);
+    auto parsed = parseDeviceNameIntoConfig(deviceName, config_with_batch, is_proxy_device(deviceName));
     auto plugin = get_plugin(parsed._deviceName);
     ov::SoPtr<ov::ICompiledModel> res;
     auto cacheManager = coreConfig.get_cache_config_for_device(plugin, parsed._config)._cacheManager;
@@ -1095,8 +1095,9 @@ std::shared_ptr<const ov::Model> ov::CoreImpl::apply_auto_batching(const std::sh
             const auto disabled = batch_mode->second.as<std::string>() == CONFIG_VALUE(NO);
             // virtual plugins like AUTO/MULTI will need the config
             // e.g. to deduce the #requests correctly
+            // proxy plugin should also keep the config
             // otherwise, no need for this config key in the rest of loading
-            if (!is_virtual_device(deviceName))
+            if (!is_virtual_device(deviceName) && !is_proxy_device(deviceName))
                 config.erase(batch_mode);
             if (disabled)
                 return model;
diff --git a/src/plugins/proxy/tests/batch_compliance_test.cpp b/src/plugins/proxy/tests/batch_compliance_test.cpp
new file mode 100644
index 00000000000..73cf894291f
--- /dev/null
+++ b/src/plugins/proxy/tests/batch_compliance_test.cpp
@@ -0,0 +1,34 @@
+// Copyright (C) 2018-2023 Intel Corporation
+// SPDX-License-Identifier: Apache-2.0
+//
+
+#include "openvino/proxy/properties.hpp"
+#include "openvino/runtime/properties.hpp"
+#include "proxy_tests.hpp"
+
+using namespace ov::proxy::tests;
+
+TEST_F(ProxyTests, can_parse_and_inherit_batch_property) {
+    register_plugin_support_reshape(core, "MOCK_DEVICE", {{ov::proxy::configuration::alias.name(), "ALIAS_MOCK"}});
+    auto available_devices = core.get_available_devices();
+    auto model = create_model_with_add();
+    auto compiled_model_default = core.compile_model(model, "MOCK_DEVICE", ov::hint::performance_mode("THROUGHPUT"));
+#ifdef ENABLE_AUTO_BATCH
+    EXPECT_NO_THROW(compiled_model_default.get_property(ov::auto_batch_timeout));  // batch enabled by default
+    EXPECT_EQ(compiled_model_default.get_property(ov::auto_batch_timeout), 1000);  // default value
+#endif
+    auto compiled_model_with_batch = core.compile_model(model,
+                                                        "MOCK_DEVICE",
+                                                        ov::hint::performance_mode("THROUGHPUT"),
+                                                        ov::hint::allow_auto_batching(true),
+                                                        ov::auto_batch_timeout(8));
+#ifdef ENABLE_AUTO_BATCH
+    EXPECT_NO_THROW(compiled_model_with_batch.get_property(ov::auto_batch_timeout));
+    EXPECT_EQ(compiled_model_with_batch.get_property(ov::auto_batch_timeout), 8);
+#endif
+    auto compiled_model_no_batch = core.compile_model(model,
+                                                      "MOCK_DEVICE",
+                                                      ov::hint::performance_mode("THROUGHPUT"),
+                                                      ov::hint::allow_auto_batching(false));
+    EXPECT_ANY_THROW(compiled_model_no_batch.get_property(ov::auto_batch_timeout));
+}
\ No newline at end of file