diff --git a/libcudacxx/test/libcudacxx/cuda/numeric/overflow.arithmetic/mul_overflow.pass.cpp b/libcudacxx/test/libcudacxx/cuda/numeric/overflow.arithmetic/mul_overflow.pass.cpp
index bd2b1ea2c17..bc0f6950275 100644
--- a/libcudacxx/test/libcudacxx/cuda/numeric/overflow.arithmetic/mul_overflow.pass.cpp
+++ b/libcudacxx/test/libcudacxx/cuda/numeric/overflow.arithmetic/mul_overflow.pass.cpp
@@ -9,6 +9,7 @@
 
 #include <cuda/numeric>
 #include <cuda/std/cassert>
+#include <cuda/std/cstdint>
 #include <cuda/std/limits>
 #include <cuda/std/tuple>
 #include <cuda/std/type_traits>
@@ -83,6 +84,7 @@ TEST_FUNC constexpr bool test_type()
   test_mul_overflow<Res, L, R>(17, 14, 238);
   test_mul_overflow<Res, L, R>(-254, 127, -32258);
   test_mul_overflow<Res, L, R>(1657, -13748, -22780436);
+  test_mul_overflow<Res, L, R>(-50000, -50000, 2500000000);
   test_mul_overflow<Res, L, R>(-2147483647, 4294967295, -9223372030412324865);
   if constexpr (cuda::std::is_unsigned_v<L> && cuda::std::is_unsigned_v<Res>)
   {
@@ -93,6 +95,44 @@ TEST_FUNC constexpr bool test_type()
       sizeof(L) >= sizeof(Res));
   }
 
+  // 5. Test T_MIN * T_MIN and T_MAX * T_MAX
+  if constexpr (sizeof(L) < sizeof(cuda::std::__cccl_uintmax_t) && sizeof(R) < sizeof(cuda::std::__cccl_uintmax_t))
+  {
+    constexpr auto __max_nbits = cuda::std::max(cuda::std::__num_bits_v<L>, cuda::std::__num_bits_v<R>);
+    using _Up = cuda::std::__make_nbit_int_t<2 * __max_nbits, cuda::std::is_signed_v<L> || cuda::std::is_signed_v<R>>;
+    test_mul_overflow<Res, L, R>(cuda::std::numeric_limits<L>::min(),
+                                 cuda::std::numeric_limits<R>::min(),
+                                 _Up{cuda::std::numeric_limits<L>::min()} * _Up{cuda::std::numeric_limits<R>::min()});
+    test_mul_overflow<Res, L, R>(cuda::std::numeric_limits<L>::max(),
+                                 cuda::std::numeric_limits<R>::max(),
+                                 _Up{cuda::std::numeric_limits<L>::max()} * _Up{cuda::std::numeric_limits<R>::max()});
+  }
+
+#if _CCCL_HAS_INT128()
+  // 6. Test __uint128_t multiplication and overflow cases
+  if constexpr (cuda::std::is_same_v<Res, __uint128_t> && cuda::std::is_same_v<R, __uint128_t>)
+  {
+    if constexpr (cuda::std::is_same_v<L, __uint128_t>)
+    {
+      test_mul_overflow<Res, L, R>(
+        __uint128_t{~0ull}, // 2^64 - 1
+        __uint128_t{1ull} << 63, // 2^63
+        (__uint128_t{0x7fffffffffffffffULL} << 64) | __uint128_t{0x8000000000000000ULL},
+        false);
+      test_mul_overflow<Res, L, R>(__uint128_t{1} << 100, __uint128_t{1} << 100, __uint128_t{0}, true);
+      test_mul_overflow<Res, L, R>(
+        cuda::std::numeric_limits<__uint128_t>::max(),
+        __uint128_t{2},
+        cuda::std::numeric_limits<__uint128_t>::max() - 1,
+        true);
+    }
+    else if constexpr (cuda::std::is_same_v<L, unsigned long long>)
+    {
+      test_mul_overflow<Res, L, R>(
+        ~0ull, __uint128_t{5} << 100, __uint128_t{0xffffffb000000000ULL} << 64 | __uint128_t{0}, true);
+    }
+  }
+#endif // _CCCL_HAS_INT128()
   return true;
 }
 
@@ -102,7 +142,7 @@ using TypeList = cuda::std::tuple<
   short,
   unsigned short,
   int,
-  unsigned long,
+  unsigned int,
   long,
   unsigned long,
   long long,