diff --git a/src/plugins/intel_npu/src/al/include/intel_npu/npuw_private_properties.hpp b/src/plugins/intel_npu/src/al/include/intel_npu/npuw_private_properties.hpp
index 5d6c6da22eb994..67dce9621bfb4e 100644
--- a/src/plugins/intel_npu/src/al/include/intel_npu/npuw_private_properties.hpp
+++ b/src/plugins/intel_npu/src/al/include/intel_npu/npuw_private_properties.hpp
@@ -30,8 +30,8 @@ static constexpr ov::Property<std::string> devices{"NPUW_DEVICES"};
  * @brief
  * Type: std::string.
  * Force the specific subgraph to specific device. The device must be present in the NPUW_DEVICES list.
- * Possible values: Comma-separated "Subgraph index:OpenVINO device name" pairs,
- * e.g. "0:CPU,1:NPU".
+ * Possible values: Comma-separated "Subgraph index:OpenVINO device name" pairs, "last" keyword can be
+ * used for last subgraph, e.g. "0:CPU,1:NPU,last:CPU".
  * Default value: empty.
  */
 static constexpr ov::Property<std::string> submodel_device{"NPUW_SUBMODEL_DEVICE"};
@@ -323,7 +323,9 @@ static constexpr ov::Property<bool> full{"NPUW_DUMP_FULL"};
  * Type: std::string.
  * Dump the specified subgraph(s) in OpenVINO IR form in the current directory.
  * Possible values: Comma-separated list of subgraph indices or "YES" for all
- * subgraphs, "NO" or just empty value to turn option off. E.g. "0,1" or "YES".
+ * subgraphs, "NO" or just empty value to turn option off. Keyword "last" can
+ * be used for dumping last subgraph without specifying it by specific index.
+ * E.g. "0,1" or "0,1,last" or "YES".
  * Default value: empty.
  */
 static constexpr ov::Property<std::string> subgraphs{"NPUW_DUMP_SUBS"};
@@ -333,7 +335,8 @@ static constexpr ov::Property<std::string> subgraphs{"NPUW_DUMP_SUBS"};
  * Type: std::string.
  * Dump subgraph on disk if a compilation failure happens.
  * Possible values: Comma-separated list of subgraph indices or "YES" for all
- * subgraphs, "NO" or just empty value to turn option off. E.g. "0,1" or "YES".
+ * subgraphs, "NO" or just empty value to turn option off. Keyword "last" can
+ * be used for dumping last subgraph. E.g. "0,1" or "0,1,last" or "YES".
  * Default value: empty.
  */
 static constexpr ov::Property<std::string> subgraphs_on_fail{"NPUW_DUMP_SUBS_ON_FAIL"};
@@ -343,7 +346,8 @@ static constexpr ov::Property<std::string> subgraphs_on_fail{"NPUW_DUMP_SUBS_ON_
  * Type: std::string.
  * Dump input & output tensors for subgraph(s).
  * Possible values: Comma-separated list of subgraph indices or "YES" for all
- * subgraphs, "NO" or just empty value to turn option off. E.g. "0,1" or "YES".
+ * subgraphs, "NO" or just empty value to turn option off. Keyword "last" can
+ * be used for last subgraph. E.g. "0,1" or "0,1,last" or "YES".
  * Default value: empty.
  */
 static constexpr ov::Property<std::string> inputs_outputs{"NPUW_DUMP_IO"};
diff --git a/src/plugins/intel_npu/src/plugin/npuw/base_sync_infer_request.cpp b/src/plugins/intel_npu/src/plugin/npuw/base_sync_infer_request.cpp
index 02422e8c72a199..216b1a35b4315c 100644
--- a/src/plugins/intel_npu/src/plugin/npuw/base_sync_infer_request.cpp
+++ b/src/plugins/intel_npu/src/plugin/npuw/base_sync_infer_request.cpp
@@ -211,7 +211,8 @@ void ov::npuw::IBaseInferRequest::infer() {
 
 void ov::npuw::IBaseInferRequest::dump_input_tensors(std::size_t idx) {
     const std::string dump_ios_opt = m_npuw_model->m_cfg.get<::intel_npu::NPUW_DUMP_IO>();
-    if (!ov::npuw::util::is_set(idx, dump_ios_opt)) {
+    const std::size_t end_idx = m_npuw_model->m_compiled_submodels.size();
+    if (!ov::npuw::util::is_set(idx, dump_ios_opt, end_idx)) {
         return;
     }
 
@@ -288,7 +289,8 @@ void ov::npuw::IBaseInferRequest::dump_input_tensors(std::size_t idx) {
 
 void ov::npuw::IBaseInferRequest::dump_output_tensors(std::size_t idx) {
     const std::string dump_ios_opt = m_npuw_model->m_cfg.get<::intel_npu::NPUW_DUMP_IO>();
-    if (!ov::npuw::util::is_set(idx, dump_ios_opt)) {
+    const std::size_t end_idx = m_npuw_model->m_compiled_submodels.size();
+    if (!ov::npuw::util::is_set(idx, dump_ios_opt, end_idx)) {
         return;
     }
 
diff --git a/src/plugins/intel_npu/src/plugin/npuw/compiled_model.cpp b/src/plugins/intel_npu/src/plugin/npuw/compiled_model.cpp
index 69d68e020b887b..4bf92588e0ebbd 100644
--- a/src/plugins/intel_npu/src/plugin/npuw/compiled_model.cpp
+++ b/src/plugins/intel_npu/src/plugin/npuw/compiled_model.cpp
@@ -5,6 +5,7 @@
 
 #include <iostream>
 #include <memory>
+#include <string>
 
 #include "accuracy/comparator.hpp"
 #include "intel_npu/npu_private_properties.hpp"
@@ -256,6 +257,7 @@ ov::npuw::CompiledModel::CompiledModel(const std::shared_ptr<ov::Model>& model,
     // - dump the subgraphs, if necessary
     std::map<std::string, std::size_t> compiledFunctions;
     m_compiled_submodels.resize(orderedSubgraphs.size());
+    const std::size_t end_sub_idx = orderedSubgraphs.size();
 
     const std::string dump_sub_opt = m_cfg.get<::intel_npu::NPUW_DUMP_SUBS>();
 
@@ -323,7 +325,7 @@ ov::npuw::CompiledModel::CompiledModel(const std::shared_ptr<ov::Model>& model,
             fill_empty_tensor_names(m_compiled_submodels[real_id].model);
         }
 
-        if (ov::npuw::util::is_set(id, dump_sub_opt)) {
+        if (ov::npuw::util::is_set(id, dump_sub_opt, end_sub_idx)) {
             LOG_INFO("Dumping Subgraph[" << id << "]");
             LOG_BLOCK();
             if (real_id != id) {
@@ -340,7 +342,14 @@ ov::npuw::CompiledModel::CompiledModel(const std::shared_ptr<ov::Model>& model,
     }      // for(orderedSubgraphs)
 
     std::map<std::size_t, std::string> forced_sub_devices{};
-    const std::string fsd_opt = m_cfg.get<::intel_npu::NPUW_SUBMODEL_DEVICE>();
+    std::string fsd_opt = m_cfg.get<::intel_npu::NPUW_SUBMODEL_DEVICE>();
+    // Change "last" keyword to tail subgraph number
+    std::size_t last_pos = fsd_opt.find("last");
+    if (last_pos != std::string::npos) {
+        fsd_opt.erase(last_pos, 4);
+        fsd_opt.insert(last_pos, std::to_string(end_sub_idx));
+    }
+
     forced_sub_devices = ::intel_npu ::OptionParser<std::map<std::size_t, std::string>>::parse(fsd_opt);
 
     // Exclude optimized out subgraphs from compilation target beforehand - otherwise we might get head and repeated
@@ -694,8 +703,9 @@ ov::SoPtr<ov::ICompiledModel> ov::npuw::CompiledModel::compile_submodel(const st
 
 void ov::npuw::CompiledModel::dump_on_fail(std::size_t id, const std::string& device_to_try, const char* extra) {
     const std::string dof_opt = m_cfg.get<::intel_npu::NPUW_DUMP_SUBS_ON_FAIL>();
+    const std::size_t end_idx = m_compiled_submodels.size();
 
-    if (ov::npuw::util::is_set(id, dof_opt)) {
+    if (ov::npuw::util::is_set(id, dof_opt, end_idx)) {
         ov::npuw::dump_failure(m_compiled_submodels[id].model, device_to_try, extra);
     }
 }
diff --git a/src/plugins/intel_npu/src/plugin/npuw/util.cpp b/src/plugins/intel_npu/src/plugin/npuw/util.cpp
index dfef4defb48c04..e9deb34ee2ded7 100644
--- a/src/plugins/intel_npu/src/plugin/npuw/util.cpp
+++ b/src/plugins/intel_npu/src/plugin/npuw/util.cpp
@@ -17,7 +17,7 @@
 #include "openvino/runtime/make_tensor.hpp"  // get_tensor_impl
 #include "util_xarch.hpp"
 
-bool ov::npuw::util::is_set(const std::size_t sub_idx, const std::string& opt) {
+bool ov::npuw::util::is_set(const std::size_t sub_idx, const std::string& opt, const std::size_t end_idx) {
     if (opt.empty() || opt == "NO") {
         return false;
     }
@@ -25,12 +25,20 @@ bool ov::npuw::util::is_set(const std::size_t sub_idx, const std::string& opt) {
         return true;
     }
 
+    std::string str(opt);
+    std::size_t last_pos = str.find("last");
+    if (last_pos != std::string::npos) {
+        str.erase(last_pos, 4);
+        if (end_idx != SIZE_MAX && sub_idx == end_idx - 1) {
+            return true;
+        }
+    }
+
     std::vector<std::size_t> sub_inds{};
-    sub_inds = ::intel_npu ::OptionParser<std::vector<std::size_t>>::parse(opt);
+    sub_inds = ::intel_npu ::OptionParser<std::vector<std::size_t>>::parse(str);
     if (std::find(sub_inds.begin(), sub_inds.end(), sub_idx) != sub_inds.end()) {
         return true;
     }
-
     return false;
 }
 
diff --git a/src/plugins/intel_npu/src/plugin/npuw/util.hpp b/src/plugins/intel_npu/src/plugin/npuw/util.hpp
index a826d00e032977..c7b54ee89fd010 100644
--- a/src/plugins/intel_npu/src/plugin/npuw/util.hpp
+++ b/src/plugins/intel_npu/src/plugin/npuw/util.hpp
@@ -15,7 +15,7 @@ namespace ov {
 namespace npuw {
 namespace util {
 
-bool is_set(const std::size_t sub_idx, const std::string& opt);
+bool is_set(const std::size_t sub_idx, const std::string& opt, const std::size_t end_idx = SIZE_MAX);
 
 // Every great project has its own string class...
 // NB: Newer C++ standards would allow to use string views or smt