diff --git a/stl/inc/vector b/stl/inc/vector
index 17902275c58..5a18de75a39 100644
--- a/stl/inc/vector
+++ b/stl/inc/vector
@@ -2882,6 +2882,65 @@ namespace pmr {
     using vector = _STD vector<_Ty, polymorphic_allocator<_Ty>>;
 } // namespace pmr
 #endif // _HAS_CXX17
+
+#if _HAS_IF_CONSTEXPR
+// VARIABLE TEMPLATE _Is_vb_iterator
+template <class _Alloc, bool _RequiresMutable>
+_INLINE_VAR constexpr bool _Is_vb_iterator<_Vb_iterator<_Alloc>, _RequiresMutable> = true;
+
+template <class _Alloc>
+_INLINE_VAR constexpr bool _Is_vb_iterator<_Vb_const_iterator<_Alloc>, false> = true;
+
+template <class _FwdIt, class _Ty>
+_CONSTEXPR20 void _Fill_vbool(_FwdIt _First, _FwdIt _Last, const _Ty& _Val) {
+    // Set [_First, _Last) to _Val
+    if (_First == _Last) {
+        return;
+    }
+
+    _Vbase* _VbFirst      = const_cast<_Vbase*>(_First._Myptr);
+    _Vbase* const _VbLast = const_cast<_Vbase*>(_Last._Myptr);
+
+    const auto _FirstSourceMask = static_cast<_Vbase>(-1) << _First._Myoff;
+    const auto _FirstDestMask   = ~_FirstSourceMask;
+    const auto _FillVal         = static_cast<_Vbase>(_Val ? -1 : 0);
+
+    if (_VbFirst == _VbLast) {
+        // We already excluded _First == _Last, so here _Last._Myoff > 0 and the shift is safe
+        const auto _LastSourceMask = static_cast<_Vbase>(-1) >> (_VBITS - _Last._Myoff);
+        const auto _LastDestMask   = ~_LastSourceMask;
+        const auto _SourceMask     = _FirstSourceMask & _LastSourceMask;
+        const auto _DestMask       = _FirstDestMask | _LastDestMask;
+        *_VbFirst                  = (*_VbFirst & _DestMask) | (_FillVal & _SourceMask);
+        return;
+    }
+
+    *_VbFirst = (*_VbFirst & _FirstDestMask) | (_FillVal & _FirstSourceMask);
+    ++_VbFirst;
+
+#ifdef __cpp_lib_is_constant_evaluated
+    if (_STD is_constant_evaluated()) {
+        for (; _VbFirst != _VbLast; ++_VbFirst) {
+            *_VbFirst = _FillVal;
+        }
+    } else
+#endif // __cpp_lib_is_constant_evaluated
+    {
+        const auto _VbFirst_ch = reinterpret_cast<const char*>(_VbFirst);
+        const auto _VbLast_ch  = reinterpret_cast<const char*>(_VbLast);
+        const auto _Count_ch   = static_cast<size_t>(_VbLast_ch - _VbFirst_ch);
+        const auto _ValChar    = static_cast<unsigned char>(_Val ? -1 : 0);
+        _CSTD memset(_VbFirst, _ValChar, _Count_ch);
+        _VbFirst = _VbLast;
+    }
+
+    if (_Last._Myoff != 0) {
+        const auto _LastSourceMask = static_cast<_Vbase>(-1) >> (_VBITS - _Last._Myoff);
+        const auto _LastDestMask   = ~_LastSourceMask;
+        *_VbFirst                  = (*_VbFirst & _LastDestMask) | (_FillVal & _LastSourceMask);
+    }
+}
+#endif // _HAS_IF_CONSTEXPR
 _STD_END
 
 #pragma pop_macro("new")
diff --git a/stl/inc/xutility b/stl/inc/xutility
index b22d75d57ef..0bebdd19d2b 100644
--- a/stl/inc/xutility
+++ b/stl/inc/xutility
@@ -4302,6 +4302,10 @@ _OutIt _Copy_memmove(move_iterator<_InIt> _First, move_iterator<_InIt> _Last, _O
 }
 
 #if _HAS_IF_CONSTEXPR
+// VARIABLE TEMPLATE _Is_vb_iterator
+template <class _It, bool _RequiresMutable = false>
+_INLINE_VAR constexpr bool _Is_vb_iterator = false;
+
 template <class _InIt, class _OutIt>
 _CONSTEXPR20 _OutIt _Copy_unchecked(_InIt _First, _InIt _Last, _OutIt _Dest) {
     // copy [_First, _Last) to [_Dest, ...)
@@ -4717,20 +4721,24 @@ template <class _FwdIt, class _Ty>
 _CONSTEXPR20 void fill(const _FwdIt _First, const _FwdIt _Last, const _Ty& _Val) {
     // copy _Val through [_First, _Last)
     _Adl_verify_range(_First, _Last);
-    auto _UFirst      = _Get_unwrapped(_First);
-    const auto _ULast = _Get_unwrapped(_Last);
-    if constexpr (_Fill_memset_is_safe<decltype(_UFirst), _Ty>) {
+    if constexpr (_Is_vb_iterator<_FwdIt, true>) {
+        _Fill_vbool(_First, _Last, _Val);
+    } else {
+        auto _UFirst      = _Get_unwrapped(_First);
+        const auto _ULast = _Get_unwrapped(_Last);
+        if constexpr (_Fill_memset_is_safe<decltype(_UFirst), _Ty>) {
 #ifdef __cpp_lib_is_constant_evaluated
-        if (!_STD is_constant_evaluated())
+            if (!_STD is_constant_evaluated())
 #endif // __cpp_lib_is_constant_evaluated
-        {
-            _CSTD memset(_UFirst, static_cast<unsigned char>(_Val), static_cast<size_t>(_ULast - _UFirst));
-            return;
+            {
+                _CSTD memset(_UFirst, static_cast<unsigned char>(_Val), static_cast<size_t>(_ULast - _UFirst));
+                return;
+            }
         }
-    }
 
-    for (; _UFirst != _ULast; ++_UFirst) {
-        *_UFirst = _Val;
+        for (; _UFirst != _ULast; ++_UFirst) {
+            *_UFirst = _Val;
+        }
     }
 }
 #else // ^^^ _HAS_IF_CONSTEXPR // !_HAS_IF_CONSTEXPR vvv
@@ -4773,26 +4781,31 @@ _CONSTEXPR20 _OutIt fill_n(_OutIt _Dest, const _Diff _Count_raw, const _Ty& _Val
     // copy _Val _Count times through [_Dest, ...)
     _Algorithm_int_t<_Diff> _Count = _Count_raw;
     if (0 < _Count) {
-        auto _UDest = _Get_unwrapped_n(_Dest, _Count);
-        if constexpr (_Fill_memset_is_safe<decltype(_UDest), _Ty>) {
+        if constexpr (_Is_vb_iterator<_OutIt, true>) {
+            const auto _Last = _Dest + static_cast<typename _OutIt::difference_type>(_Count);
+            _Fill_vbool(_Dest, _Last, _Val);
+            return _Last;
+        } else {
+            auto _UDest = _Get_unwrapped_n(_Dest, _Count);
+            if constexpr (_Fill_memset_is_safe<decltype(_UDest), _Ty>) {
 #ifdef __cpp_lib_is_constant_evaluated
-            if (!_STD is_constant_evaluated())
+                if (!_STD is_constant_evaluated())
 #endif // __cpp_lib_is_constant_evaluated
-            {
-                _CSTD memset(_UDest, static_cast<unsigned char>(_Val), static_cast<size_t>(_Count));
-                _UDest += _Count;
-                _Seek_wrapped(_Dest, _UDest);
-                return _Dest;
+                {
+                    _CSTD memset(_UDest, static_cast<unsigned char>(_Val), static_cast<size_t>(_Count));
+                    _UDest += _Count;
+                    _Seek_wrapped(_Dest, _UDest);
+                    return _Dest;
+                }
             }
-        }
 
-        for (; 0 < _Count; --_Count, (void) ++_UDest) {
-            *_UDest = _Val;
-        }
+            for (; 0 < _Count; --_Count, (void) ++_UDest) {
+                *_UDest = _Val;
+            }
 
-        _Seek_wrapped(_Dest, _UDest);
+            _Seek_wrapped(_Dest, _UDest);
+        }
     }
-
     return _Dest;
 }
 
diff --git a/tests/std/test.lst b/tests/std/test.lst
index 04b21077e19..d639d5e2c81 100644
--- a/tests/std/test.lst
+++ b/tests/std/test.lst
@@ -157,6 +157,7 @@ tests\Dev11_1158803_regex_thread_safety
 tests\Dev11_1180290_filesystem_error_code
 tests\GH_000457_system_error_message
 tests\GH_000545_include_compare
+tests\GH_000625_vector_bool_optimization
 tests\GH_000685_condition_variable_any
 tests\GH_000690_overaligned_function
 tests\GH_000890_pow_template
diff --git a/tests/std/tests/GH_000625_vector_bool_optimization/env.lst b/tests/std/tests/GH_000625_vector_bool_optimization/env.lst
new file mode 100644
index 00000000000..19f025bd0e6
--- /dev/null
+++ b/tests/std/tests/GH_000625_vector_bool_optimization/env.lst
@@ -0,0 +1,4 @@
+# Copyright (c) Microsoft Corporation.
+# SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+
+RUNALL_INCLUDE ..\usual_matrix.lst
diff --git a/tests/std/tests/GH_000625_vector_bool_optimization/test.cpp b/tests/std/tests/GH_000625_vector_bool_optimization/test.cpp
new file mode 100644
index 00000000000..3f26088c87b
--- /dev/null
+++ b/tests/std/tests/GH_000625_vector_bool_optimization/test.cpp
@@ -0,0 +1,101 @@
+// Copyright (c) Microsoft Corporation.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+
+#include <algorithm>
+#include <assert.h>
+#include <vector>
+
+using namespace std;
+
+constexpr int blockSize = 32;
+static_assert(blockSize == _VBITS, "Invalid block size");
+
+void test_fill_helper(const size_t length) {
+    // No offset
+    {
+        vector<bool> result_true(length, true);
+        result_true.resize(length + 3, false);
+        vector<bool> dest_true(length + 3, false);
+        fill(dest_true.begin(), prev(dest_true.end(), 3), true);
+        assert(dest_true == result_true);
+
+        vector<bool> result_false(length, false);
+        result_false.resize(length + 3, true);
+        vector<bool> dest_false(length + 3, true);
+        fill(dest_false.begin(), prev(dest_false.end(), 3), false);
+        assert(dest_false == result_false);
+
+        vector<bool> result_true_n(length, true);
+        result_true_n.resize(length + 3, false);
+        vector<bool> dest_true_n(length + 3, false);
+        const auto res_fill_n = fill_n(dest_true_n.begin(), length, true);
+        assert(dest_true_n == result_true_n);
+        assert(res_fill_n == prev(dest_true_n.end(), 3));
+
+        vector<bool> result_false_n(length, false);
+        result_false_n.resize(length + 3, true);
+        vector<bool> dest_false_n(length + 3, true);
+        fill_n(dest_false_n.begin(), length, false);
+        assert(dest_false_n == result_false_n);
+    }
+
+    // With offset
+    {
+        vector<bool> result_true(length, true);
+        result_true.resize(length + 3, false);
+        result_true.insert(result_true.begin(), false);
+        vector<bool> dest_true(length + 4, false);
+        fill(next(dest_true.begin()), prev(dest_true.end(), 3), true);
+        assert(dest_true == result_true);
+
+        vector<bool> result_false(length, false);
+        result_false.resize(length + 3, true);
+        result_false.insert(result_false.begin(), true);
+        vector<bool> dest_false(length + 4, true);
+        fill(next(dest_false.begin()), prev(dest_false.end(), 3), false);
+        assert(dest_false == result_false);
+
+        vector<bool> result_true_n(length, true);
+        result_true_n.resize(length + 3, false);
+        result_true_n.insert(result_true_n.begin(), false);
+        vector<bool> dest_true_n(length + 4, false);
+        const auto res_fill_n = fill_n(next(dest_true_n.begin()), length, true);
+        assert(dest_true_n == result_true_n);
+        assert(res_fill_n == prev(dest_true_n.end(), 3));
+
+        vector<bool> result_false_n(length, false);
+        result_false_n.resize(length + 3, true);
+        result_false_n.insert(result_false_n.begin(), true);
+        vector<bool> dest_false_n(length + 4, true);
+        fill_n(next(dest_false_n.begin()), length, false);
+        assert(dest_false_n == result_false_n);
+    }
+}
+
+bool test_fill() {
+    // Empty
+    test_fill_helper(0);
+
+    // One block, ends within block
+    test_fill_helper(15);
+
+    // One block, ends at block boundary
+    test_fill_helper(blockSize);
+
+    // Multiple blocks, no memset, within block
+    test_fill_helper(blockSize + 11);
+
+    // Multiple blocks, no memset, ends at block boundary
+    test_fill_helper(2 * blockSize);
+
+    // Multiple blocks, with memset, within block
+    test_fill_helper(3 * blockSize + 5);
+
+    // Multiple blocks, with memset, ends at block boundary
+    test_fill_helper(4 * blockSize);
+    return true;
+}
+
+int main() {
+    test_fill();
+}