diff --git a/src/plugins/intel_gpu/src/graph/impls/cpu/detection_output.cpp b/src/plugins/intel_gpu/src/graph/impls/cpu/detection_output.cpp
index 7fbc7f38801..3fb57d21cd8 100644
--- a/src/plugins/intel_gpu/src/graph/impls/cpu/detection_output.cpp
+++ b/src/plugins/intel_gpu/src/graph/impls/cpu/detection_output.cpp
@@ -282,6 +282,12 @@ public:
         auto out_ptr = lock.begin();
 
         const auto& args = instance.argument;
+
+        auto confidence_layout = instance.confidence_memory()->get_layout();
+        auto priors_layout = instance.prior_box_memory()->get_layout();
+
+        const int num_of_priors = priors_layout.spatial(1) / args->prior_info_size;
+        const int num_classes = (args->num_classes == -1) ? confidence_layout.feature() / num_of_priors : args->num_classes;
         // Per image -> For each label: Pair (score, prior index)
         std::vector<std::map<int, std::vector<std::pair<float, int>>>> final_detections;
         for (int image = 0; image < num_of_images; ++image) {
@@ -290,7 +296,7 @@ public:
             std::map<int, std::vector<int>> indices;
             int num_det = 0;
             if (nms_type == NMSType::CAFFE) {
-                for (int cls = 0; cls < static_cast<int>(args->num_classes); ++cls) {
+                for (int cls = 0; cls < num_classes; ++cls) {
                     if (static_cast<int>(cls) == args->background_label_id) {
                         conf_per_image[cls].clear();
                         continue;  // Skip background class.
@@ -522,9 +528,7 @@ public:
     template <typename dtype>
     void extract_confidences_per_image_caffe(stream& stream, const detection_output_inst& instance,
                                              std::vector<std::vector<std::vector<std::pair<float, int>>>>& confidences,
-                                             const int num_of_priors) {
-        const int num_classes = instance.argument->num_classes;
-
+                                             const int num_of_priors, const int num_classes) {
         const int num_of_images = static_cast<int>(confidences.size());
         auto input_confidence = instance.confidence_memory();
         const float confidence_threshold = instance.argument->confidence_threshold;
@@ -616,9 +620,8 @@ public:
     template <typename dtype>
     void extract_confidences_per_image_mxnet(stream& stream, const detection_output_inst& instance,
                                              std::vector<std::vector<std::vector<std::pair<float, int>>>>& confidences,
-                                             const int num_of_priors,
+                                             const int num_of_priors, const int num_classes,
                                              std::vector<std::vector<std::pair<float, std::pair<int, int>>>>& scoreIndexPairs) {
-        const int num_classes = instance.argument->num_classes;
         const int background_label_id = instance.argument->background_label_id;
         const int num_of_images = static_cast<int>(confidences.size());
         auto input_confidence = instance.confidence_memory();
@@ -750,11 +753,13 @@ public:
 
         const auto& args = instance.argument;
 
+        auto confidence_layout = instance.confidence_memory()->get_layout();
         auto priors_layout = instance.prior_box_memory()->get_layout();
 
         const int num_of_images = static_cast<int>(bboxes.size());
         const int num_of_priors = priors_layout.spatial(1) / args->prior_info_size;
-        const int num_loc_classes = args->share_location ? 1 : args->num_classes;
+        const int num_classes = (args->num_classes == -1) ? confidence_layout.feature() / num_of_priors : args->num_classes;
+        const int num_loc_classes = args->share_location ? 1 : num_classes;
 
         // Extract locations per image.
         std::vector<std::vector<std::vector<bounding_box>>> locations(
@@ -812,9 +817,9 @@ public:
         }
         // Extract confidences per image.
         if (nms_type == NMSType::CAFFE) {
-            extract_confidences_per_image_caffe<dtype>(stream, instance, confidences, num_of_priors);
+            extract_confidences_per_image_caffe<dtype>(stream, instance, confidences, num_of_priors, num_classes);
         } else {
-            extract_confidences_per_image_mxnet<dtype>(stream, instance, confidences, num_of_priors, scoreIndexPairs);
+            extract_confidences_per_image_mxnet<dtype>(stream, instance, confidences, num_of_priors, num_classes, scoreIndexPairs);
         }
     }
 
diff --git a/src/plugins/intel_gpu/tests/functional/single_layer_tests/dynamic/detection_output.cpp b/src/plugins/intel_gpu/tests/functional/single_layer_tests/dynamic/detection_output.cpp
index 5cb201d48ca..cba4a9d88b3 100644
--- a/src/plugins/intel_gpu/tests/functional/single_layer_tests/dynamic/detection_output.cpp
+++ b/src/plugins/intel_gpu/tests/functional/single_layer_tests/dynamic/detection_output.cpp
@@ -197,9 +197,24 @@ protected:
 
         auto params = ngraph::builder::makeDynamicParams(ngraph::element::f32, inputDynamicShapes);
         auto paramOuts = ngraph::helpers::convert2OutputVector(ngraph::helpers::castOps2Nodes<ngraph::opset3::Parameter>(params));
-        auto detOut = ngraph::builder::makeDetectionOutput(paramOuts, attrs);
-        ngraph::ResultVector results{std::make_shared<ngraph::opset3::Result>(detOut)};
-        function = std::make_shared<ngraph::Function>(results, params, "DetectionOutputDynamic");
+
+        if (attrs.num_classes == -1) {
+            std::shared_ptr<ov::op::v8::DetectionOutput> detOut;
+
+            if (paramOuts.size() == 3)
+                detOut = std::make_shared<ov::op::v8::DetectionOutput>(paramOuts[0], paramOuts[1], paramOuts[2], attrs);
+            else if (paramOuts.size() == 5)
+                detOut = std::make_shared<ov::op::v8::DetectionOutput>(paramOuts[0], paramOuts[1], paramOuts[2], paramOuts[3], paramOuts[4], attrs);
+            else
+                throw std::runtime_error("DetectionOutput layer supports only 3 or 5 inputs");
+
+            ngraph::ResultVector results{std::make_shared<ngraph::opset3::Result>(detOut)};
+            function = std::make_shared<ngraph::Function>(results, params, "DetectionOutputDynamic");
+        } else {
+            auto detOut = ngraph::builder::makeDetectionOutput(paramOuts, attrs);
+            ngraph::ResultVector results{std::make_shared<ngraph::opset3::Result>(detOut)};
+            function = std::make_shared<ngraph::Function>(results, params, "DetectionOutputDynamic");
+        }
     }
 
 private:
@@ -242,7 +257,7 @@ TEST_P(DetectionOutputLayerGPUTest, CompareWithRefs) {
 
 namespace {
 
-const int numClasses = 11;
+const std::vector<int> numClasses = {11, -1};
 const int backgroundLabelId = 0;
 const std::vector<int> topK = {75};
 const std::vector<std::vector<int>> keepTopK = { {50}, {100} };
@@ -256,7 +271,7 @@ const float objectnessScore = 0.4f;
 const std::vector<size_t> numberBatch = {1, 2};
 
 const auto commonAttributes = ::testing::Combine(
-    ::testing::Values(numClasses),
+    ::testing::Values(numClasses[0]),
     ::testing::Values(backgroundLabelId),
     ::testing::ValuesIn(topK),
     ::testing::ValuesIn(keepTopK),
@@ -268,6 +283,18 @@ const auto commonAttributes = ::testing::Combine(
     ::testing::ValuesIn(decreaseLabelId)
 );
 
+const auto commonAttributes_v8 = ::testing::Combine(
+    ::testing::Values(numClasses[1]),
+    ::testing::Values(backgroundLabelId),
+    ::testing::Values(topK[0]),
+    ::testing::Values(keepTopK[0]),
+    ::testing::ValuesIn(codeType),
+    ::testing::Values(nmsThreshold),
+    ::testing::Values(confidenceThreshold),
+    ::testing::Values(clipAfterNms[0]),
+    ::testing::Values(clipBeforeNms[0]),
+    ::testing::Values(decreaseLabelId[0])
+);
 /* =============== 3 inputs cases =============== */
 
 const std::vector<ParamsWhichSizeDependsDynamic> specificParams3InDynamic = {
@@ -362,9 +389,21 @@ const auto params3InputsDynamic = ::testing::Combine(
         ::testing::Values(ov::test::utils::DEVICE_GPU)
 );
 
+const auto params3InputsDynamic_v8 = ::testing::Combine(
+        commonAttributes_v8,
+        ::testing::Values(specificParams3InDynamic[0]),
+        ::testing::ValuesIn(numberBatch),
+        ::testing::Values(objectnessScore),
+        ::testing::Values(true),
+        ::testing::Values(ov::test::utils::DEVICE_GPU)
+);
+
 INSTANTIATE_TEST_SUITE_P(smoke_GPUDetectionOutputDynamic3In, DetectionOutputLayerGPUTest,
                          params3InputsDynamic,
                          DetectionOutputLayerGPUTest::getTestCaseName);
 
+INSTANTIATE_TEST_SUITE_P(smoke_GPUDetectionOutputV8Dynamic3In, DetectionOutputLayerGPUTest,
+                         params3InputsDynamic_v8,
+                         DetectionOutputLayerGPUTest::getTestCaseName);
 } // namespace
 } // namespace GPULayerTestsDefinitions