diff --git a/src/core/shape_inference/include/one_hot_shape_inference.hpp b/src/core/shape_inference/include/one_hot_shape_inference.hpp
index f6f944de456680..ece6634cf638bc 100644
--- a/src/core/shape_inference/include/one_hot_shape_inference.hpp
+++ b/src/core/shape_inference/include/one_hot_shape_inference.hpp
@@ -67,7 +67,7 @@ std::vector<TRShape> shape_infer(const OneHot* op,
         auto depth_as_shape =
             get_input_const_data_as_shape<TRShape>(op, 1, ta, util::GetNotNegative<typename DimType::value_type>(op));
 
-        if (depth_as_shape && depth_as_shape->size() == 1 && (*depth_as_shape)[0].get_length() > 0) {
+        if (depth_as_shape && depth_as_shape->size() == 1) {
             result_shape.insert(result_shape.begin() + axis, (*depth_as_shape)[0]);
         } else {
             result_shape.insert(result_shape.begin() + axis, DimType());
diff --git a/src/plugins/intel_gpu/include/intel_gpu/primitives/one_hot.hpp b/src/plugins/intel_gpu/include/intel_gpu/primitives/one_hot.hpp
index caec53c857378c..d47a128c62c434 100644
--- a/src/plugins/intel_gpu/include/intel_gpu/primitives/one_hot.hpp
+++ b/src/plugins/intel_gpu/include/intel_gpu/primitives/one_hot.hpp
@@ -63,7 +63,7 @@ struct one_hot : public primitive_base<one_hot> {
             const int64_t& one_hot_axis,
             const float& on_value = 1.0f,
             const float& off_value = 0.0f)
-        : primitive_base(id, {input, input_depth})
+        : primitive_base(id, {input, input_depth}, 1, {optional_data_type{output_dt}})
         , shape(shape)
         , one_hot_axis(one_hot_axis)
         , on_value(on_value)
diff --git a/src/plugins/intel_gpu/src/graph/one_hot.cpp b/src/plugins/intel_gpu/src/graph/one_hot.cpp
index fe810a9bd8bf5c..e9613f360fd380 100644
--- a/src/plugins/intel_gpu/src/graph/one_hot.cpp
+++ b/src/plugins/intel_gpu/src/graph/one_hot.cpp
@@ -66,14 +66,15 @@ std::vector<layout> one_hot_inst::calc_output_layouts(const one_hot_node& /*node
     };
 
     int64_t depth = desc->depth;
-
-    auto depth_tensor = ov::Tensor(ov::element::i64, ov::Shape{1}, static_cast<void*>(&depth));
-    std::unordered_map<size_t, ov::Tensor> const_data = {
-        {1, depth_tensor}
-    };
-
     auto& memory_deps = impl_param.memory_deps;
-    if (memory_deps.count(1) > 0) {
+
+    std::unordered_map<size_t, ov::Tensor> const_data = {};
+    if (depth != 0) {
+        auto depth_tensor = ov::Tensor(ov::element::i64, ov::Shape{1}, static_cast<void*>(&depth));
+        const_data = {
+            {1, depth_tensor}
+        };
+    } else if (memory_deps.count(1) > 0) {
         auto depth_mem = memory_deps.at(1);
 
         cldnn::mem_lock<uint8_t, mem_lock_type::read> depth_lock(depth_mem, impl_param.get_stream());