diff --git a/extension/tensor/tensor_ptr.h b/extension/tensor/tensor_ptr.h
index 1ec73882573..a6b377b2ef1 100644
--- a/extension/tensor/tensor_ptr.h
+++ b/extension/tensor/tensor_ptr.h
@@ -15,28 +15,6 @@ namespace executorch {
 namespace extension {
 
 #ifndef USE_ATEN_LIB
-namespace internal {
-/**
- * Custom deleter for TensorPtr that ensures proper management of the associated
- * TensorImplPtr.
- *
- * Since Tensor does not own its TensorImpl, this deleter manages the
- * TensorImplPtr lifecycle, ensuring dynamic metadata (sizes, dim_order,
- * strides) is released appropriately when the Tensor is destroyed.
- */
-struct TensorPtrDeleter final {
-  TensorImplPtr tensor_impl;
-
-  void operator()(exec_aten::Tensor* pointer) {
-    // Release all resources immediately since the data held by the
-    // TensorPtrDeleter is tied to the managed object, not the smart pointer
-    // itself. We need to free this memory when the object is destroyed, not
-    // when the smart pointer (and deleter) are eventually destroyed or reset.
-    tensor_impl.reset();
-    delete pointer;
-  }
-};
-} // namespace internal
 
 /**
  * A smart pointer for managing the lifecycle of a Tensor.
@@ -47,8 +25,62 @@ struct TensorPtrDeleter final {
  * by design. It ensures that the underlying TensorImpl can be safely shared
  * among tensors as needed.
  */
-using TensorPtr =
-    std::unique_ptr<exec_aten::Tensor, internal::TensorPtrDeleter>;
+class TensorPtr : private std::unique_ptr<exec_aten::Tensor> {
+ public:
+  using unique_ptr::element_type;
+  using unique_ptr::get;
+  using unique_ptr::pointer;
+  using unique_ptr::operator bool;
+  using unique_ptr::operator*;
+  using unique_ptr::operator->;
+
+  constexpr TensorPtr() = default;
+  constexpr TensorPtr(std::nullptr_t) {}
+  ~TensorPtr() = default;
+  TensorPtr(TensorPtr&& rhs) noexcept = default;
+  TensorPtr& operator=(TensorPtr&& rhs) noexcept = default;
+
+  TensorPtr(TensorImplPtr p)
+      : unique_ptr(std::make_unique<exec_aten::Tensor>(p.get())),
+        tensor_impl(std::move(p)) {}
+
+  // release() does not make sense as it separates the Tensor from the
+  // TensorImplPtr. using unique_ptr::release;
+
+  void reset() {
+    unique_ptr::reset();
+    tensor_impl.reset();
+  }
+
+  void swap(TensorPtr& other) {
+    unique_ptr::swap(static_cast<unique_ptr&>(other));
+    tensor_impl.swap(other.tensor_impl);
+  }
+
+  bool operator==(const TensorPtr& rhs) const {
+    // No need to check tensor_impl; if the Tensor pointers are equal
+    // then the tensor_impls must also be since this is a unique
+    // pointer.
+    return static_cast<const unique_ptr&>(*this) ==
+        static_cast<const unique_ptr&>(rhs);
+  }
+
+  bool operator!=(const TensorPtr& rhs) const {
+    return !(*this == rhs);
+  }
+
+  bool operator==(std::nullptr_t) const {
+    return static_cast<const unique_ptr&>(*this) == nullptr;
+  }
+
+  bool operator!=(std::nullptr_t) const {
+    return !(*this == nullptr);
+  }
+
+ private:
+  friend TensorPtr make_tensor_ptr(const TensorPtr& tensor);
+  TensorImplPtr tensor_impl;
+};
 #else
 /**
  * A smart pointer type for managing the lifecycle of a Tensor.
@@ -74,9 +106,7 @@ using TensorPtr = std::unique_ptr<exec_aten::Tensor>;
  */
 inline TensorPtr make_tensor_ptr(TensorImplPtr tensor_impl) {
 #ifndef USE_ATEN_LIB
-  auto tensor = std::make_unique<exec_aten::Tensor>(tensor_impl.get());
-  return TensorPtr(
-      tensor.release(), internal::TensorPtrDeleter{std::move(tensor_impl)});
+  return TensorPtr(std::move(tensor_impl));
 #else
   return std::make_unique<exec_aten::Tensor>(std::move(tensor_impl));
 #endif // USE_ATEN_LIB
@@ -96,7 +126,7 @@ inline TensorPtr make_tensor_ptr(TensorImplPtr tensor_impl) {
  */
 inline TensorPtr make_tensor_ptr(const TensorPtr& tensor) {
 #ifndef USE_ATEN_LIB
-  return make_tensor_ptr(tensor.get_deleter().tensor_impl);
+  return make_tensor_ptr(tensor.tensor_impl);
 #else
   return make_tensor_ptr(tensor->getIntrusivePtr());
 #endif // USE_ATEN_LIB