diff --git a/libcxx/docs/Status/RangesAlgorithms.csv b/libcxx/docs/Status/RangesAlgorithms.csv
--- a/libcxx/docs/Status/RangesAlgorithms.csv
+++ b/libcxx/docs/Status/RangesAlgorithms.csv
@@ -64,7 +64,7 @@
 Merge,set_difference,Hui Xie,`D128983 <https://llvm.org/D128983>`,✅
 Merge,set_intersection,Hui Xie,`D129233 <https://llvm.org/D129233>`,✅
 Merge,set_symmetric_difference,Hui Xie,`D129520 <https://llvm.org/D129520>`,✅
-Merge,set_union,Hui Xie,n/a,Not started
+Merge,set_union,Hui Xie,`D129657 <https://llvm.org/D129657>`,✅
 Permutation,remove,Nikolas Klauser,`D128618 <https://llvm.org/D128618>`_,✅
 Permutation,remove_if,Nikolas Klauser,`D128618 <https://llvm.org/D128618>`_,✅
 Permutation,reverse,Nikolas Klauser,`D125752 <https://llvm.org/D125752>`_,✅
diff --git a/libcxx/include/__algorithm/ranges_set_union.h b/libcxx/include/__algorithm/ranges_set_union.h
--- a/libcxx/include/__algorithm/ranges_set_union.h
+++ b/libcxx/include/__algorithm/ranges_set_union.h
@@ -42,34 +42,68 @@
 namespace __set_union {
 
 struct __fn {
-
-  template <input_iterator _InIter1, sentinel_for<_InIter1> _Sent1,
-            input_iterator _InIter2, sentinel_for<_InIter2> _Sent2,
-            weakly_incrementable _OutIter, class _Comp = ranges::less,
-            class _Proj1 = identity, class _Proj2 = identity>
-  requires mergeable<_InIter1, _InIter2, _OutIter, _Comp, _Proj1, _Proj2>
-  _LIBCPP_HIDE_FROM_ABI constexpr
-  set_union_result<_InIter1, _InIter2, _OutIter>
-  operator()(_InIter1 __first1, _Sent1 __last1, _InIter2 __first2, _Sent2 __last2, _OutIter __result, _Comp __comp = {},
-            _Proj1 __proj1 = {}, _Proj2 __proj2 = {}) const {
-    // TODO: implement
-    (void)__first1; (void)__last1; (void)__first2; (void)__last2; (void)__result; (void)__comp; (void)__proj1;
-    (void)__proj2;
-    return {};
+  template <
+      input_iterator _InIter1,
+      sentinel_for<_InIter1> _Sent1,
+      input_iterator _InIter2,
+      sentinel_for<_InIter2> _Sent2,
+      weakly_incrementable _OutIter,
+      class _Comp  = ranges::less,
+      class _Proj1 = identity,
+      class _Proj2 = identity>
+    requires mergeable<_InIter1, _InIter2, _OutIter, _Comp, _Proj1, _Proj2>
+  _LIBCPP_HIDE_FROM_ABI constexpr set_union_result<_InIter1, _InIter2, _OutIter> operator()(
+      _InIter1 __first1,
+      _Sent1 __last1,
+      _InIter2 __first2,
+      _Sent2 __last2,
+      _OutIter __result,
+      _Comp __comp   = {},
+      _Proj1 __proj1 = {},
+      _Proj2 __proj2 = {}) const {
+    auto __ret = std::__set_union(
+        std::move(__first1),
+        std::move(__last1),
+        std::move(__first2),
+        std::move(__last2),
+        std::move(__result),
+        ranges::__make_projected_comp(__comp, __proj1, __proj2));
+    return {std::move(__ret.__in1_), std::move(__ret.__in2_), std::move(__ret.__out_)};
   }
 
-  template <input_range _Range1, input_range _Range2, weakly_incrementable _OutIter,
-            class _Comp = ranges::less, class _Proj1 = identity, class _Proj2 = identity>
-  requires mergeable<iterator_t<_Range1>, iterator_t<_Range2>, _OutIter, _Comp, _Proj1, _Proj2>
-  _LIBCPP_HIDE_FROM_ABI constexpr
-  set_union_result<borrowed_iterator_t<_Range1>, borrowed_iterator_t<_Range2>, _OutIter>
-  operator()(_Range1&& __range1, _Range2&& __range2, _OutIter __result, _Comp __comp = {},
-            _Proj1 __proj1 = {}, _Proj2 __proj2 = {}) const {
-    // TODO: implement
-    (void)__range1; (void)__range2; (void)__result; (void)__comp; (void)__proj1; (void)__proj2;
-    return {};
+  template <
+      input_range _Range1,
+      input_range _Range2,
+      weakly_incrementable _OutIter,
+      class _Comp  = ranges::less,
+      class _Proj1 = identity,
+      class _Proj2 = identity>
+    requires mergeable<
+        iterator_t<_Range1>,
+        iterator_t<_Range2>,
+        _OutIter,
+        _Comp,
+        _Proj1,
+        _Proj2> 
+  _LIBCPP_HIDE_FROM_ABI constexpr set_union_result<borrowed_iterator_t<_Range1>,
+                                                   borrowed_iterator_t<_Range2>,
+                                                   _OutIter>
+    operator()(
+        _Range1&& __range1,
+        _Range2&& __range2,
+        _OutIter __result,
+        _Comp __comp   = {},
+        _Proj1 __proj1 = {},
+        _Proj2 __proj2 = {}) const {
+    auto __ret = std::__set_union(
+        ranges::begin(__range1),
+        ranges::end(__range1),
+        ranges::begin(__range2),
+        ranges::end(__range2),
+        std::move(__result),
+        ranges::__make_projected_comp(__comp, __proj1, __proj2));
+    return {std::move(__ret.__in1_), std::move(__ret.__in2_), std::move(__ret.__out_)};
   }
-
 };
 
 } // namespace __set_union
diff --git a/libcxx/include/__algorithm/set_union.h b/libcxx/include/__algorithm/set_union.h
--- a/libcxx/include/__algorithm/set_union.h
+++ b/libcxx/include/__algorithm/set_union.h
@@ -14,6 +14,7 @@
 #include <__algorithm/copy.h>
 #include <__config>
 #include <__iterator/iterator_traits.h>
+#include <__utility/move.h>
 
 #if !defined(_LIBCPP_HAS_NO_PRAGMA_SYSTEM_HEADER)
 #  pragma GCC system_header
@@ -21,50 +22,77 @@
 
 _LIBCPP_BEGIN_NAMESPACE_STD
 
-template <class _Compare, class _InputIterator1, class _InputIterator2, class _OutputIterator>
-_LIBCPP_CONSTEXPR_AFTER_CXX17 _OutputIterator
-__set_union(_InputIterator1 __first1, _InputIterator1 __last1,
-            _InputIterator2 __first2, _InputIterator2 __last2, _OutputIterator __result, _Compare __comp)
-{
-    for (; __first1 != __last1; ++__result)
-    {
-        if (__first2 == __last2)
-            return _VSTD::copy(__first1, __last1, __result);
-        if (__comp(*__first2, *__first1))
-        {
-            *__result = *__first2;
-            ++__first2;
-        }
-        else
-        {
-            if (!__comp(*__first1, *__first2))
-                ++__first2;
-            *__result = *__first1;
-            ++__first1;
-        }
+template <class _InIter1, class _InIter2, class _OutIter>
+struct __set_union_result {
+  _InIter1 __in1_;
+  _InIter2 __in2_;
+  _OutIter __out_;
+
+  // need a constructor as C++03 aggregate init is hard
+  _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_AFTER_CXX17
+  __set_union_result(_InIter1&& __in_iter1, _InIter2&& __in_iter2, _OutIter&& __out_iter)
+      : __in1_(std::move(__in_iter1)), __in2_(std::move(__in_iter2)), __out_(std::move(__out_iter)) {}
+};
+
+template <class _Compare, class _InIter1, class _Sent1, class _InIter2, class _Sent2, class _OutIter>
+_LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_AFTER_CXX17 __set_union_result<_InIter1, _InIter2, _OutIter> __set_union(
+    _InIter1 __first1, _Sent1 __last1, _InIter2 __first2, _Sent2 __last2, _OutIter __result, _Compare&& __comp) {
+  for (; __first1 != __last1; ++__result) {
+    if (__first2 == __last2) {
+      auto __ret1 = std::__copy_impl(std::move(__first1), std::move(__last1), std::move(__result));
+      return __set_union_result<_InIter1, _InIter2, _OutIter>(
+          std::move(__ret1.first), std::move(__first2), std::move((__ret1.second)));
+    }
+    if (__comp(*__first2, *__first1)) {
+      *__result = *__first2;
+      ++__first2;
+    } else {
+      if (!__comp(*__first1, *__first2)) {
+        ++__first2;
+      }
+      *__result = *__first1;
+      ++__first1;
     }
-    return _VSTD::copy(__first2, __last2, __result);
+  }
+  auto __ret2 = std::__copy_impl(std::move(__first2), std::move(__last2), std::move(__result));
+  return __set_union_result<_InIter1, _InIter2, _OutIter>(
+      std::move(__first1), std::move(__ret2.first), std::move((__ret2.second)));
 }
 
 template <class _InputIterator1, class _InputIterator2, class _OutputIterator, class _Compare>
-inline _LIBCPP_INLINE_VISIBILITY _LIBCPP_CONSTEXPR_AFTER_CXX17
-_OutputIterator
-set_union(_InputIterator1 __first1, _InputIterator1 __last1,
-          _InputIterator2 __first2, _InputIterator2 __last2, _OutputIterator __result, _Compare __comp)
-{
-    typedef typename __comp_ref_type<_Compare>::type _Comp_ref;
-    return _VSTD::__set_union<_Comp_ref>(__first1, __last1, __first2, __last2, __result, __comp);
+_LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_AFTER_CXX17 _OutputIterator set_union(
+    _InputIterator1 __first1,
+    _InputIterator1 __last1,
+    _InputIterator2 __first2,
+    _InputIterator2 __last2,
+    _OutputIterator __result,
+    _Compare __comp) {
+  typedef typename __comp_ref_type<_Compare>::type _Comp_ref;
+  return std::__set_union<_Comp_ref>(
+             std::move(__first1),
+             std::move(__last1),
+             std::move(__first2),
+             std::move(__last2),
+             std::move(__result),
+             __comp)
+      .__out_;
 }
 
 template <class _InputIterator1, class _InputIterator2, class _OutputIterator>
-inline _LIBCPP_INLINE_VISIBILITY _LIBCPP_CONSTEXPR_AFTER_CXX17
-_OutputIterator
-set_union(_InputIterator1 __first1, _InputIterator1 __last1,
-          _InputIterator2 __first2, _InputIterator2 __last2, _OutputIterator __result)
-{
-    return _VSTD::set_union(__first1, __last1, __first2, __last2, __result,
-                          __less<typename iterator_traits<_InputIterator1>::value_type,
-                                 typename iterator_traits<_InputIterator2>::value_type>());
+_LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_AFTER_CXX17 _OutputIterator set_union(
+    _InputIterator1 __first1,
+    _InputIterator1 __last1,
+    _InputIterator2 __first2,
+    _InputIterator2 __last2,
+    _OutputIterator __result) {
+  return std::set_union(
+      std::move(__first1),
+      std::move(__last1),
+      std::move(__first2),
+      std::move(__last2),
+      std::move(__result),
+      __less<typename iterator_traits<_InputIterator1>::value_type,
+             typename iterator_traits<_InputIterator2>::value_type>());
 }
 
 _LIBCPP_END_NAMESPACE_STD
diff --git a/libcxx/include/algorithm b/libcxx/include/algorithm
--- a/libcxx/include/algorithm
+++ b/libcxx/include/algorithm
@@ -712,6 +712,23 @@
       set_symmetric_difference(R1&& r1, R2&& r2, O result, Comp comp = {},
                                Proj1 proj1 = {}, Proj2 proj2 = {});                                 // since C++20
   
+  template<class I1, class I2, class O>
+    using set_union_result = in_in_out_result<I1, I2, O>;                                           // since C++20
+
+  template<input_iterator I1, sentinel_for<I1> S1, input_iterator I2, sentinel_for<I2> S2,
+           weakly_incrementable O, class Comp = ranges::less,
+           class Proj1 = identity, class Proj2 = identity>
+    requires mergeable<I1, I2, O, Comp, Proj1, Proj2>
+    constexpr set_union_result<I1, I2, O>
+      set_union(I1 first1, S1 last1, I2 first2, S2 last2, O result, Comp comp = {},
+                Proj1 proj1 = {}, Proj2 proj2 = {});                                                // since C++20
+
+  template<input_range R1, input_range R2, weakly_incrementable O,
+           class Comp = ranges::less, class Proj1 = identity, class Proj2 = identity>
+    requires mergeable<iterator_t<R1>, iterator_t<R2>, O, Comp, Proj1, Proj2>
+    constexpr set_union_result<borrowed_iterator_t<R1>, borrowed_iterator_t<R2>, O>
+      set_union(R1&& r1, R2&& r2, O result, Comp comp = {},
+                Proj1 proj1 = {}, Proj2 proj2 = {});                                                // since C++20
 }
 
     constexpr bool     // constexpr in C++20
@@ -1493,6 +1510,7 @@
 #include <__algorithm/ranges_set_difference.h>
 #include <__algorithm/ranges_set_intersection.h>
 #include <__algorithm/ranges_set_symmetric_difference.h>
+#include <__algorithm/ranges_set_union.h>
 #include <__algorithm/ranges_sort.h>
 #include <__algorithm/ranges_sort_heap.h>
 #include <__algorithm/ranges_stable_sort.h>
diff --git a/libcxx/test/libcxx/algorithms/ranges_robust_against_copying_comparators.pass.cpp b/libcxx/test/libcxx/algorithms/ranges_robust_against_copying_comparators.pass.cpp
--- a/libcxx/test/libcxx/algorithms/ranges_robust_against_copying_comparators.pass.cpp
+++ b/libcxx/test/libcxx/algorithms/ranges_robust_against_copying_comparators.pass.cpp
@@ -205,8 +205,8 @@
     (void)std::ranges::set_intersection(a, b, first2, Less(&copies)); assert(copies == 0);
     (void)std::ranges::set_symmetric_difference(first, mid, mid, last, first2, Less(&copies)); assert(copies == 0);
     (void)std::ranges::set_symmetric_difference(a, b, first2, Less(&copies)); assert(copies == 0);
-    //(void)std::ranges::set_union(first, mid, mid, last, first2, Less(&copies)); assert(copies == 0);
-    //(void)std::ranges::set_union(a, b, first2, Less(&copies)); assert(copies == 0);
+    (void)std::ranges::set_union(first, mid, mid, last, first2, Less(&copies)); assert(copies == 0);
+    (void)std::ranges::set_union(a, b, first2, Less(&copies)); assert(copies == 0);
     (void)std::ranges::sort(first, last, Less(&copies)); assert(copies == 0);
     (void)std::ranges::sort(a, Less(&copies)); assert(copies == 0);
     (void)std::ranges::sort_heap(first, last, Less(&copies)); assert(copies == 0);
diff --git a/libcxx/test/libcxx/algorithms/ranges_robust_against_copying_projections.pass.cpp b/libcxx/test/libcxx/algorithms/ranges_robust_against_copying_projections.pass.cpp
--- a/libcxx/test/libcxx/algorithms/ranges_robust_against_copying_projections.pass.cpp
+++ b/libcxx/test/libcxx/algorithms/ranges_robust_against_copying_projections.pass.cpp
@@ -196,8 +196,8 @@
     (void)std::ranges::set_intersection(a, b, first2, Less(), Proj(&copies), Proj(&copies)); assert(copies == 0);
     (void)std::ranges::set_symmetric_difference(first, mid, mid, last, first2, Less(), Proj(&copies), Proj(&copies)); assert(copies == 0);
     (void)std::ranges::set_symmetric_difference(a, b, first2, Less(), Proj(&copies), Proj(&copies)); assert(copies == 0);
-    //(void)std::ranges::set_union(first, mid, mid, last, first2, Less(), Proj(&copies), Proj(&copies)); assert(copies == 0);
-    //(void)std::ranges::set_union(a, b, first2, Less(), Proj(&copies), Proj(&copies)); assert(copies == 0);
+    (void)std::ranges::set_union(first, mid, mid, last, first2, Less(), Proj(&copies), Proj(&copies)); assert(copies == 0);
+    (void)std::ranges::set_union(a, b, first2, Less(), Proj(&copies), Proj(&copies)); assert(copies == 0);
     (void)std::ranges::sort(first, last, Less(), Proj(&copies)); assert(copies == 0);
     (void)std::ranges::sort(a, Less(), Proj(&copies)); assert(copies == 0);
     (void)std::ranges::sort_heap(first, last, Less(), Proj(&copies)); assert(copies == 0);
diff --git a/libcxx/test/std/algorithms/alg.sorting/alg.set.operations/set.union/ranges_set_union.pass.cpp b/libcxx/test/std/algorithms/alg.sorting/alg.set.operations/set.union/ranges_set_union.pass.cpp
--- a/libcxx/test/std/algorithms/alg.sorting/alg.set.operations/set.union/ranges_set_union.pass.cpp
+++ b/libcxx/test/std/algorithms/alg.sorting/alg.set.operations/set.union/ranges_set_union.pass.cpp
@@ -31,17 +31,464 @@
 #include <concepts>
 #include <functional>
 #include <ranges>
+#include <utility>
 
 #include "almost_satisfies_types.h"
+#include "MoveOnly.h"
 #include "test_iterators.h"
+#include "../../sortable_helpers.h"
 
-// TODO: SFINAE tests.
+// Test iterator overload's constraints:
+// =====================================
+template <class InIter1 = int*, class Sent1 = int*, class InIter2 = int*, class Sent2 = int*, class OutIter = int*>
+concept HasSetUnionIter =
+    requires(InIter1&& inIter1, InIter2&& inIter2, OutIter&& outIter, Sent1&& sent1, Sent2&& sent2) {
+      std::ranges::set_union(
+          std::forward<InIter1>(inIter1),
+          std::forward<Sent1>(sent1),
+          std::forward<InIter2>(inIter2),
+          std::forward<Sent2>(sent2),
+          std::forward<OutIter>(outIter));
+    };
+
+static_assert(HasSetUnionIter<int*, int*, int*, int*, int*>);
+
+// !std::input_iterator<I1>
+static_assert(!HasSetUnionIter<InputIteratorNotDerivedFrom>);
+
+// !std::sentinel_for<S1, I1>
+static_assert(!HasSetUnionIter<int*, SentinelForNotSemiregular>);
+
+// !std::input_iterator<I2>
+static_assert(!HasSetUnionIter<int*, int*, InputIteratorNotDerivedFrom>);
+
+// !std::sentinel_for<S2, I2>
+static_assert(!HasSetUnionIter<int*, int*, int*, SentinelForNotSemiregular>);
+
+// !std::weakly_incrementable<O>
+static_assert(!HasSetUnionIter<int*, int*, int*, int*, WeaklyIncrementableNotMovable>);
+
+// !std::mergeable<I1, I2, O, Comp, Proj1, Proj2>
+static_assert(!HasSetUnionIter<MoveOnly*, MoveOnly*, MoveOnly*, MoveOnly*, MoveOnly*>);
+
+// Test range overload's constraints:
+// =====================================
+
+template <class Range1, class Range2, class OutIter>
+concept HasSetUnionRange =
+    requires(Range1&& range1, Range2&& range2, OutIter&& outIter) {
+      std::ranges::set_union(
+          std::forward<Range1>(range1), std::forward<Range2>(range2), std::forward<OutIter>(outIter));
+    };
+
+template <class T>
+using R = UncheckedRange<T>;
+
+static_assert(HasSetUnionRange<R<int*>, R<int*>, int*>);
+
+// !std::input_range<R2>
+static_assert(!HasSetUnionRange<R<InputIteratorNotDerivedFrom>, R<int*>, int*>);
+
+// !std::input_range<R2>
+static_assert(!HasSetUnionRange<R<int*>, R<InputIteratorNotDerivedFrom>, int*>);
+
+// !std::weakly_incrementable<O>
+static_assert(!HasSetUnionRange<R<int*>, R<int*>, WeaklyIncrementableNotMovable >);
+
+// !std::mergeable<iterator_t<R1>, iterator_t<R2>, O, Comp, Proj1, Proj2>
+static_assert(!HasSetUnionRange<R<MoveOnly*>, R<MoveOnly*>, MoveOnly*>);
+
+using std::ranges::set_union_result;
+
+template <class In1, class In2, class Out, std::size_t N1, std::size_t N2, std::size_t N3>
+constexpr void testSetUnionImpl(std::array<int, N1> in1, std::array<int, N2> in2, std::array<int, N3> expected) {
+  // TODO: std::ranges::set_union calls std::ranges::copy
+  // std::ranges::copy(contiguous_iterator<int*>, sentinel_wrapper<contiguous_iterator<int*>>, contiguous_iterator<int*>) doesn't seem to work.
+  // It seems that std::ranges::copy calls std::copy, which unwraps contiguous_iterator<int*> into int*,
+  // and then it failed because there is no == between int* and sentinel_wrapper<contiguous_iterator<int*>>
+  using Sent1 = std::conditional_t<std::contiguous_iterator<In1>, In1, sentinel_wrapper<In1>>;
+  using Sent2 = std::conditional_t<std::contiguous_iterator<In2>, In2, sentinel_wrapper<In2>>;
+
+  // iterator overload
+  {
+    std::array<int, N3> out;
+    std::same_as<set_union_result<In1, In2, Out>> decltype(auto) result = std::ranges::set_union(
+        In1{in1.data()},
+        Sent1{In1{in1.data() + in1.size()}},
+        In2{in2.data()},
+        Sent2{In2{in2.data() + in2.size()}},
+        Out{out.data()});
+    assert(std::ranges::equal(out, expected));
+
+    assert(base(result.in1) == in1.data() + in1.size());
+    assert(base(result.in2) == in2.data() + in2.size());
+    assert(base(result.out) == out.data() + out.size());
+  }
+
+  // range overload
+  {
+    std::array<int, N3> out;
+    std::ranges::subrange r1{In1{in1.data()}, Sent1{In1{in1.data() + in1.size()}}};
+    std::ranges::subrange r2{In2{in2.data()}, Sent2{In2{in2.data() + in2.size()}}};
+    std::same_as<set_union_result<In1, In2, Out>> decltype(auto) result =
+        std::ranges::set_union(r1, r2, Out{out.data()});
+    assert(std::ranges::equal(out, expected));
+
+    assert(base(result.in1) == in1.data() + in1.size());
+    assert(base(result.in2) == in2.data() + in2.size());
+    assert(base(result.out) == out.data() + out.size());
+  }
+}
+
+template <class In1, class In2, class Out>
+constexpr void testImpl() {
+  // range 1 shorter than range2
+  {
+    std::array in1{0, 1, 5, 6, 9, 10};
+    std::array in2{3, 6, 7, 9, 13, 15, 100};
+    std::array expected{0, 1, 3, 5, 6, 7, 9, 10, 13, 15, 100};
+    testSetUnionImpl<In1, In2, Out>(in1, in2, expected);
+  }
+  // range 2 shorter than range 1
+  {
+    std::array in1{2, 6, 8, 12, 15, 16};
+    std::array in2{0, 2, 8};
+    std::array expected{0, 2, 6, 8, 12, 15, 16};
+    testSetUnionImpl<In1, In2, Out>(in1, in2, expected);
+  }
+
+  // range 1 and range 2 has the same length but different elements
+  {
+    std::array in1{2, 6, 8, 12, 15, 16};
+    std::array in2{0, 2, 8, 15, 17, 19};
+    std::array expected{0, 2, 6, 8, 12, 15, 16, 17, 19};
+    testSetUnionImpl<In1, In2, Out>(in1, in2, expected);
+  }
+
+  // range 1 == range 2
+  {
+    std::array in1{0, 1, 2};
+    std::array in2{0, 1, 2};
+    std::array expected{0, 1, 2};
+    testSetUnionImpl<In1, In2, Out>(in1, in2, expected);
+  }
+
+  // range 1 is super set of range 2
+  {
+    std::array in1{8, 8, 10, 12, 13};
+    std::array in2{8, 10};
+    std::array expected{8, 8, 10, 12, 13};
+    testSetUnionImpl<In1, In2, Out>(in1, in2, expected);
+  }
+
+  // range 2 is super set of range 1
+  {
+    std::array in1{0, 1, 1};
+    std::array in2{0, 1, 1, 2, 5};
+    std::array expected{0, 1, 1, 2, 5};
+    testSetUnionImpl<In1, In2, Out>(in1, in2, expected);
+  }
+
+  // range 1 and range 2 have no elements in common
+  {
+    std::array in1{7, 7, 9, 12};
+    std::array in2{1, 5, 5, 8, 10};
+    std::array expected{1, 5, 5, 7, 7, 8, 9, 10, 12};
+    testSetUnionImpl<In1, In2, Out>(in1, in2, expected);
+  }
+
+  // range 1 and range 2 have duplicated equal elements
+  {
+    std::array in1{7, 7, 9, 12};
+    std::array in2{7, 7, 7, 13};
+    std::array expected{7, 7, 7, 9, 12, 13};
+    testSetUnionImpl<In1, In2, Out>(in1, in2, expected);
+  }
+
+  // range 1 is empty
+  {
+    std::array<int, 0> in1{};
+    std::array in2{3, 4, 5};
+    std::array expected{3, 4, 5};
+    testSetUnionImpl<In1, In2, Out>(in1, in2, expected);
+  }
+
+  // range 2 is empty
+  {
+    std::array in1{3, 4, 5};
+    std::array<int, 0> in2{};
+    std::array expected{3, 4, 5};
+    testSetUnionImpl<In1, In2, Out>(in1, in2, expected);
+  }
+
+  // both ranges are empty
+  {
+    std::array<int, 0> in1{};
+    std::array<int, 0> in2{};
+    std::array<int, 0> expected{};
+    testSetUnionImpl<In1, In2, Out>(in1, in2, expected);
+  }
+
+  // check that ranges::dangling is returned for non-borrowed_range
+  {
+    std::array r1{3, 6, 7, 9};
+    int r2[] = {2, 3, 4, 5, 6};
+    std::array<int, 7> out;
+    std::same_as<set_union_result<std::ranges::dangling, int*, int*>> decltype(auto) result =
+        std::ranges::set_union(NonBorrowedRange<In1>{r1.data(), r1.size()}, r2, out.data());
+    assert(base(result.in2) == r2 + 5);
+    assert(base(result.out) == out.data() + out.size());
+    assert(std::ranges::equal(out, std::array{2, 3, 4, 5, 6, 7, 9}));
+  }
+}
+
+template <class InIter2, class OutIter>
+constexpr void withAllPermutationsOfInIter1() {
+  // C++17 InputIterator may or may not satisfy std::input_iterator
+  testImpl<cpp20_input_iterator<int*>, InIter2, OutIter>();
+  testImpl<forward_iterator<int*>, InIter2, OutIter>();
+  testImpl<bidirectional_iterator<int*>, InIter2, OutIter>();
+  testImpl<random_access_iterator<int*>, InIter2, OutIter>();
+  testImpl<contiguous_iterator<int*>, InIter2, OutIter>();
+}
+
+template <class OutIter>
+constexpr bool withAllPermutationsOfInIter1AndInIter2() {
+  withAllPermutationsOfInIter1<cpp20_input_iterator<int*>, OutIter>();
+  withAllPermutationsOfInIter1<forward_iterator<int*>, OutIter>();
+  withAllPermutationsOfInIter1<bidirectional_iterator<int*>, OutIter>();
+  withAllPermutationsOfInIter1<random_access_iterator<int*>, OutIter>();
+  withAllPermutationsOfInIter1<contiguous_iterator<int*>, OutIter>();
+  return true;
+}
+
+constexpr void runAllIteratorPermutationsTests() {
+  withAllPermutationsOfInIter1AndInIter2<cpp20_output_iterator<int*>>();
+  withAllPermutationsOfInIter1AndInIter2<cpp20_input_iterator<int*>>();
+  withAllPermutationsOfInIter1AndInIter2<forward_iterator<int*>>();
+  withAllPermutationsOfInIter1AndInIter2<bidirectional_iterator<int*>>();
+  withAllPermutationsOfInIter1AndInIter2<random_access_iterator<int*>>();
+  withAllPermutationsOfInIter1AndInIter2<contiguous_iterator<int*>>();
+
+  static_assert(withAllPermutationsOfInIter1AndInIter2<cpp20_output_iterator<int*>>());
+  static_assert(withAllPermutationsOfInIter1AndInIter2<cpp20_input_iterator<int*>>());
+  static_assert(withAllPermutationsOfInIter1AndInIter2<forward_iterator<int*>>());
+  static_assert(withAllPermutationsOfInIter1AndInIter2<bidirectional_iterator<int*>>());
+  static_assert(withAllPermutationsOfInIter1AndInIter2<random_access_iterator<int*>>());
+  static_assert(withAllPermutationsOfInIter1AndInIter2<contiguous_iterator<int*>>());
+}
 
 constexpr bool test() {
-  // TODO: main tests.
-  // TODO: A custom comparator works.
-  // TODO: A custom projection works.
+  // check that every element is copied exactly once
+  {
+    std::array<TracedCopy, 5> r1{3, 5, 8, 15, 16};
+    std::array<TracedCopy, 3> r2{1, 3, 8};
 
+    // iterator overload
+    {
+      std::array<TracedCopy, 6> out;
+      auto result = std::ranges::set_union(r1.begin(), r1.end(), r2.begin(), r2.end(), out.data());
+
+      assert(result.in1 == r1.end());
+      assert(result.in2 == r2.end());
+      assert(result.out == out.end());
+      assert(std::ranges::equal(out, std::array<TracedCopy, 6>{1, 3, 5, 8, 15, 16}));
+
+      assert(std::ranges::all_of(out, &TracedCopy::copiedOnce));
+    }
+
+    // range overload
+    {
+      std::array<TracedCopy, 6> out;
+      auto result = std::ranges::set_union(r1, r2, out.data());
+
+      assert(result.in1 == r1.end());
+      assert(result.in2 == r2.end());
+      assert(result.out == out.end());
+      assert(std::ranges::equal(out, std::array<TracedCopy, 6>{1, 3, 5, 8, 15, 16}));
+
+      assert(std::ranges::all_of(out, &TracedCopy::copiedOnce));
+    }
+  }
+
+  struct IntAndOrder {
+    int data;
+    int order;
+
+    constexpr auto operator==(const IntAndOrder& o) const { return data == o.data; }
+    constexpr auto operator<=>(const IntAndOrder& o) const { return data <=> o.data; }
+  };
+
+  // Stable ([algorithm.stable]). If [first1, last1) contains m elements that are
+  // equivalent to each other and [first2, last2) contains n elements that are
+  // equivalent to them, then all m elements from the first range are copied to the
+  // output range, in order, and then the final max(n−m,0) elements from the second
+  // range are copied to the output range, in order.
+  {
+    std::array<IntAndOrder, 3> r1{{{0, 0}, {0, 1}, {0, 2}}};
+    std::array<IntAndOrder, 5> r2{{{0, 3}, {0, 4}, {0, 5}, {0, 6}, {0, 7}}};
+
+    // iterator overload
+    {
+      std::array<IntAndOrder, 5> out;
+      std::ranges::set_union(r1.begin(), r1.end(), r2.begin(), r2.end(), out.data());
+
+      assert(std::ranges::equal(out, std::array{0, 0, 0, 0, 0}, {}, &IntAndOrder::data));
+      assert(std::ranges::equal(out, std::array{0, 1, 2, 6, 7}, {}, &IntAndOrder::order));
+    }
+
+    // range overload
+    {
+      std::array<IntAndOrder, 5> out;
+      std::ranges::set_union(r1, r2, out.data());
+
+      assert(std::ranges::equal(out, std::array{0, 0, 0, 0, 0}, {}, &IntAndOrder::data));
+      assert(std::ranges::equal(out, std::array{0, 1, 2, 6, 7}, {}, &IntAndOrder::order));
+    }
+  }
+
+  struct Data {
+    int data;
+  };
+
+  // Test custom comparator
+  {
+    std::array r1{Data{4}, Data{8}, Data{12}};
+    std::array r2{Data{8}, Data{9}};
+
+    // iterator overload
+    {
+      std::array<Data, 4> out;
+      auto result = std::ranges::set_union(
+          r1.begin(), r1.end(), r2.begin(), r2.end(), out.data(), [](const Data& x, const Data& y) {
+            return x.data < y.data;
+          });
+
+      assert(std::ranges::equal(out, std::array{4, 8, 9, 12}, {}, &Data::data));
+
+      assert(result.in1 == r1.end());
+      assert(result.in2 == r2.end());
+      assert(result.out == out.end());
+    }
+
+    // range overload
+    {
+      std::array<Data, 4> out;
+      auto result = std::ranges::set_union(r1, r2, out.data(), [](const Data& x, const Data& y) {
+        return x.data < y.data;
+      });
+
+      assert(std::ranges::equal(out, std::array{4, 8, 9, 12}, {}, &Data::data));
+
+      assert(result.in1 == r1.end());
+      assert(result.in2 == r2.end());
+      assert(result.out == out.end());
+    }
+  }
+
+  // Test Projection
+  {
+    std::array r1{Data{1}, Data{3}, Data{5}};
+    std::array r2{Data{2}, Data{3}, Data{5}};
+
+    const auto proj = [](const Data& d) { return d.data; };
+
+    // iterator overload
+    {
+      std::array<Data, 4> out;
+      auto result = std::ranges::set_union(
+          r1.begin(), r1.end(), r2.begin(), r2.end(), out.data(), std::ranges::less{}, proj, proj);
+
+      assert(std::ranges::equal(out, std::array{1, 2, 3, 5}, {}, &Data::data));
+
+      assert(result.in1 == r1.end());
+      assert(result.in2 == r2.end());
+      assert(result.out == out.end());
+    }
+
+    // range overload
+    {
+      std::array<Data, 4> out;
+      auto result = std::ranges::set_union(r1, r2, out.data(), std::ranges::less{}, proj, proj);
+
+      assert(std::ranges::equal(out, std::array{1, 2, 3, 5}, {}, &Data::data));
+
+      assert(result.in1 == r1.end());
+      assert(result.in2 == r2.end());
+      assert(result.out == out.end());
+    }
+  }
+
+  // Complexity: At most 2 * ((last1 - first1) + (last2 - first2)) - 1 comparisons and applications of each projection.
+  {
+    struct CompProjs {
+      std::size_t numberOfComp  = 0;
+      std::size_t numberOfProj1 = 0;
+      std::size_t numberOfProj2 = 0;
+
+      constexpr auto comp() {
+        return [this](int x, int y) {
+          ++numberOfComp;
+          return x < y;
+        };
+      }
+
+      constexpr auto proj1() {
+        return [this](const Data& d) {
+          ++numberOfProj1;
+          return d.data;
+        };
+      }
+
+      constexpr auto proj2() {
+        return [this](const Data& d) {
+          ++numberOfProj2;
+          return d.data;
+        };
+      }
+    };
+
+    std::array<Data, 3> r1{{{0}, {1}, {2}}};
+    std::array<Data, 4> r2{{{0}, {2}, {2}, {5}}};
+    std::array expected{0, 1, 2, 2, 5};
+
+    const std::size_t maxOperation = 2 * (r1.size() + r2.size()) - 1;
+
+    // iterator overload
+    {
+      std::array<Data, 5> out;
+      CompProjs compProjs{};
+
+      std::ranges::set_union(
+          r1.begin(),
+          r1.end(),
+          r2.begin(),
+          r2.end(),
+          out.data(),
+          compProjs.comp(),
+          compProjs.proj1(),
+          compProjs.proj2());
+
+      assert(std::ranges::equal(out, expected, {}, &Data::data));
+      assert(compProjs.numberOfComp < maxOperation);
+      assert(compProjs.numberOfProj1 < maxOperation);
+      assert(compProjs.numberOfProj2 < maxOperation);
+    }
+
+    // range overload
+    {
+      std::array<Data, 5> out;
+      CompProjs compProjs{};
+
+      std::ranges::set_union(r1, r2, out.data(), compProjs.comp(), compProjs.proj1(), compProjs.proj2());
+
+      assert(std::ranges::equal(out, expected, {}, &Data::data));
+      assert(compProjs.numberOfComp < maxOperation);
+      assert(compProjs.numberOfProj1 < maxOperation);
+      assert(compProjs.numberOfProj2 < maxOperation);
+    }
+  }
+  
   return true;
 }
 
@@ -49,5 +496,12 @@
   test();
   static_assert(test());
 
+  // Cannot static_assert on the entire permutation test because it exceeds the constexpr execution step limit
+  // due to the large number of combination of types of iterators (it is a 3-dimensional cartesian product)
+  // Instead of having one single static_assert that tests all the combinations, in the runAllIteratorPermutationsTests
+  // function, it has lots of smaller static_assert and each of them test 2-dimensional cartesian product which is less
+  // than the step limit.
+  runAllIteratorPermutationsTests();
+
   return 0;
 }
diff --git a/libcxx/test/std/algorithms/ranges_robust_against_nonbool_predicates.compile.pass.cpp b/libcxx/test/std/algorithms/ranges_robust_against_nonbool_predicates.compile.pass.cpp
--- a/libcxx/test/std/algorithms/ranges_robust_against_nonbool_predicates.compile.pass.cpp
+++ b/libcxx/test/std/algorithms/ranges_robust_against_nonbool_predicates.compile.pass.cpp
@@ -153,7 +153,7 @@
   in2_out_pred(std::ranges::set_difference, in, in2, out, binary_pred);
   in2_out_pred(std::ranges::set_intersection, in, in2, out, binary_pred);
   in2_out_pred(std::ranges::set_symmetric_difference, in, in2, out, binary_pred);
-  //in2_out_pred(std::ranges::set_union, in, in2, out, binary_pred);
+  in2_out_pred(std::ranges::set_union, in, in2, out, binary_pred);
   in_pred(std::ranges::remove_if, in, unary_pred);
   //in_pred(std::ranges::unique, in, binary_pred);
   //in_pred(std::ranges::partition, in, binary_pred);
diff --git a/libcxx/test/std/algorithms/ranges_robust_against_omitting_invoke.compile.pass.cpp b/libcxx/test/std/algorithms/ranges_robust_against_omitting_invoke.compile.pass.cpp
--- a/libcxx/test/std/algorithms/ranges_robust_against_omitting_invoke.compile.pass.cpp
+++ b/libcxx/test/std/algorithms/ranges_robust_against_omitting_invoke.compile.pass.cpp
@@ -194,7 +194,7 @@
   in2_out_pred(std::ranges::set_difference, in, in2, out, &Foo::binary_pred, &Bar::val, &Bar::val);
   in2_out_pred(std::ranges::set_intersection, in, in2, out, &Foo::binary_pred, &Bar::val, &Bar::val);
   in2_out_pred(std::ranges::set_symmetric_difference, in, in2, out, &Foo::binary_pred, &Bar::val, &Bar::val);
-  //in2_out_pred(std::ranges::set_union, in, in2, out, &Foo::binary_pred, &Bar::val, &Bar::val);
+  in2_out_pred(std::ranges::set_union, in, in2, out, &Foo::binary_pred, &Bar::val, &Bar::val);
   in_val(std::ranges::remove, in, x, &Bar::val);
   in_pred(std::ranges::remove_if, in, &Foo::unary_pred, &Bar::val);
   // `reverse` has neither a projection nor a predicate.
diff --git a/libcxx/test/std/library/description/conventions/customization.point.object/niebloid.compile.pass.cpp b/libcxx/test/std/library/description/conventions/customization.point.object/niebloid.compile.pass.cpp
--- a/libcxx/test/std/library/description/conventions/customization.point.object/niebloid.compile.pass.cpp
+++ b/libcxx/test/std/library/description/conventions/customization.point.object/niebloid.compile.pass.cpp
@@ -136,7 +136,7 @@
 static_assert(test(std::ranges::set_difference, a, a, a));
 static_assert(test(std::ranges::set_intersection, a, a, a));
 static_assert(test(std::ranges::set_symmetric_difference, a, a, a));
-//static_assert(test(std::ranges::set_union, a, a, a));
+static_assert(test(std::ranges::set_union, a, a, a));
 //static_assert(test(std::ranges::shuffle, a, g));
 static_assert(test(std::ranges::sort, a));
 static_assert(test(std::ranges::sort_heap, a));