diff --git a/src/VecSim/algorithms/hnsw/hnsw.h b/src/VecSim/algorithms/hnsw/hnsw.h
index 096dcc985..74c803a17 100644
--- a/src/VecSim/algorithms/hnsw/hnsw.h
+++ b/src/VecSim/algorithms/hnsw/hnsw.h
@@ -48,6 +48,11 @@ template <typename DistType>
 using candidatesLabelsMaxHeap = vecsim_stl::abstract_priority_queue<DistType, labelType>;
 using graphNodeType = pair<idType, ushort>; // represented as: (element_id, level)
 
+#if ONE_BYTE_MUTEX_AVAILABLE == 1
+using elem_mutex_t = vecsim_stl::one_byte_mutex;
+#else
+using elem_mutex_t = std::mutex;
+#endif
 ////////////////////////////////////// Auxiliary HNSW structs //////////////////////////////////////
 
 // Vectors flags (for marking a specific vector)
@@ -74,7 +79,6 @@ struct ElementMetaData {
 
     ElementMetaData(labelType label = SIZE_MAX) noexcept : label(label), flags(IN_PROCESS) {}
 };
-#pragma pack() // restore default packing
 
 struct LevelData {
     vecsim_stl::vector<idType> *incomingEdges;
@@ -94,7 +98,7 @@ struct LevelData {
 
 struct ElementGraphData {
     size_t toplevel;
-    std::mutex neighborsGuard;
+    elem_mutex_t neighborsGuard;
     LevelData *others;
     LevelData level0;
 
@@ -113,6 +117,7 @@ struct ElementGraphData {
     }
     ~ElementGraphData() = delete; // Should be destroyed using `destroyGraphData`
 };
+#pragma pack() // restore default packing
 
 //////////////////////////////////// HNSW index implementation ////////////////////////////////////
 
diff --git a/src/VecSim/utils/vecsim_stl.h b/src/VecSim/utils/vecsim_stl.h
index c5380c36a..87aa37e3c 100644
--- a/src/VecSim/utils/vecsim_stl.h
+++ b/src/VecSim/utils/vecsim_stl.h
@@ -91,4 +91,36 @@ class unordered_set
                                                                                        alloc) {}
 };
 
+#if defined(__clang__)
+#define CLANG_VERSION (__clang_major__ * 100 + __clang_minor__ * 10 + __clang_patchlevel__)
+#if (CLANG_VERSION >= 1316) // clang 13.1.6
+#define ONE_BYTE_MUTEX_AVAILABLE 1
+#endif
+#elif (__GNUC__ >= 11)
+#define ONE_BYTE_MUTEX_AVAILABLE 1
+#else
+#define ONE_BYTE_MUTEX_AVAILABLE 0
+#endif
+
+#if ONE_BYTE_MUTEX_AVAILABLE != 0
+struct one_byte_mutex {
+    void lock() {
+        if (state.exchange(locked, std::memory_order_acquire) == unlocked)
+            return;
+        while (state.exchange(sleeper, std::memory_order_acquire) != unlocked)
+            state.wait(sleeper, std::memory_order_relaxed);
+    }
+    void unlock() {
+        if (state.exchange(unlocked, std::memory_order_release) == sleeper)
+            state.notify_one();
+    }
+
+private:
+    std::atomic<uint8_t> state{unlocked};
+
+    static constexpr uint8_t unlocked = 0;
+    static constexpr uint8_t locked = 0b01;
+    static constexpr uint8_t sleeper = 0b10;
+};
+#endif
 } // namespace vecsim_stl