diff --git a/NEWS b/NEWS
index 4ba02413c..d789f406f 100644
--- a/NEWS
+++ b/NEWS
@@ -21,6 +21,8 @@ GNU coreutils NEWS                                    -*- outline -*-
 
 ** Improvements
 
+  'cut -c' is up to 10x faster in UTF-8 locales.
+
   'od' allocates half the memory when -v,--output-duplicates is specified.
 
   'tail' ensures prompt output when inotify is not being used, and when
diff --git a/src/cut.c b/src/cut.c
index 59b213f39..248e19a14 100644
--- a/src/cut.c
+++ b/src/cut.c
@@ -907,6 +907,17 @@ cut_characters_mode (FILE *stream, bool byte_mode)
 
   while (true)
     {
+      /* Skip unselected line suffixes without decoding each character.  */
+      if (idx && field_selection_exhausted (idx))
+        {
+          idx_t available = mbbuf_topup (&mbbuf);
+          char *p = mbbuf.buffer + mbbuf.offset;
+          char *end = search_bytes (p, line_delim, available);
+          mbbuf_advance (&mbbuf, end ? end - p : available);
+          if (!end && available)
+            continue;
+        }
+
       mcel_t g = mbbuf_get_char (&mbbuf);
 
       if (g.ch == line_delim)
diff --git a/tests/cut/cut.pl b/tests/cut/cut.pl
index c0c962ea0..da4dd0740 100755
--- a/tests/cut/cut.pl
+++ b/tests/cut/cut.pl
@@ -307,6 +307,10 @@ if ($mb_locale ne 'C')
        {ENV => "LC_ALL=$mb_locale"}],
       ['mb-char-5', '-c1-2', {IN=>"\xc3x\n"}, {OUT=>"\xc3x\n"},
        {ENV => "LC_ALL=$mb_locale"}],
+      # Skip unselected suffixes across input buffers and reset at line ends.
+      ['mb-char-suffix', '-c1',
+       {IN=>"\xc3\xa9" . ("\xff" x (2 * $IO_BUFSIZE)) . "\nbignored"},
+       {OUT=>"\xc3\xa9\nb\n"}, {ENV => "LC_ALL=$mb_locale"}],
       # Note mb-byte-n-1 and mb-byte-n-4 differ from coreutils-i18n patch,
       # which outputs a character if any byte is selected.
       # I.e., the i18n patch may output more bytes that the requested range.