diff --git a/CHANGELOG.md b/CHANGELOG.md
index ae81d36..f14452a 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -3,6 +3,7 @@
 * Update up Unicode 13.0.0
 * Don't use non-standard ``__uint128``
 * Add PyPy compatibility
+* Add ``decode_utf8(byte-like)``
 
 ### 1.4.7
 
diff --git a/src/_decoder.pyx b/src/_decoder.pyx
index ce2a9a6..cd3c668 100644
--- a/src/_decoder.pyx
+++ b/src/_decoder.pyx
@@ -769,15 +769,16 @@
 
 
 cdef object _decode_all(ReaderRef reader, boolean some):
-    cdef object ex
+    cdef object ex, ex2
     try:
         return _decode_all_sub(reader, some)
     except _DecoderException as ex:
-        raise (<_DecoderException> ex).cls(
+        ex2 = (<_DecoderException> ex).cls(
             (<_DecoderException> ex).msg,
             (<_DecoderException> ex).result,
             (<_DecoderException> ex).extra,
         )
+    raise ex2
 
 
 cdef object _decode_ucs1(const void *string, Py_ssize_t length,
@@ -807,12 +808,27 @@
     return _decode_all(reader, some)
 
 
+cdef object _decode_utf8(const void *string, Py_ssize_t length,
+                         Py_ssize_t maxdepth, boolean some):
+    cdef ReaderUTF8 reader = ReaderUTF8(
+        ReaderUCS(length, 0, maxdepth),
+        <const Py_UCS1*> string,
+    )
+    return _decode_all(reader, some)
+
+
 cdef object _decode_unicode(object data, Py_ssize_t maxdepth, boolean some):
     cdef Py_ssize_t length
     cdef int kind
+    cdef const char *s
 
     PyUnicode_READY(data)
 
+    if CYTHON_COMPILING_IN_PYPY:
+        length = 0
+        s = PyUnicode_AsUTF8AndSize(data, &length)
+        return _decode_utf8(s, length, maxdepth, some)
+
     length = PyUnicode_GET_LENGTH(data)
     kind = PyUnicode_KIND(data)
 
@@ -831,7 +847,10 @@
     cdef object (*decoder)(const void*, Py_ssize_t, Py_ssize_t, boolean)
     cdef Py_ssize_t length = 0
 
-    if wordlength == 1:
+    if wordlength == 0:
+        decoder = _decode_utf8
+        length = view.len // 1
+    elif wordlength == 1:
         decoder = _decode_ucs1
         length = view.len // 1
     elif wordlength == 2:
diff --git a/src/_exports.pyx b/src/_exports.pyx
index e359fac..64500a0 100644
--- a/src/_exports.pyx
+++ b/src/_exports.pyx
@@ -62,7 +62,7 @@
 
     .. code:: python
 
-        decode_buffer(b'["Hello", "world!"]') == ['Hello', 'world!']
+        decode_latin1(b'["Hello", "world!"]') == ['Hello', 'world!']
 
     Parameters
     ----------
@@ -88,6 +88,38 @@
     return decode_buffer(data, maxdepth, bool(some), 1)
 
 
+def decode_utf8(object data, object maxdepth=None, object some=False):
+    '''
+    Decodes JSON5 serialized data from a ``bytes`` object.
+
+    .. code:: python
+
+        decode_utf8(b'["H\\xe2\\x82\\xacllo", "w\\xc3\\xb6rld!"]') == ['H€llo', 'wörld!']
+
+    Parameters
+    ----------
+    data : bytes
+        JSON5 serialized data, encoded as UTF-8 or ASCII.
+    maxdepth : Optional[int]
+        see `decode(...) <pyjson5.decode_>`_
+    some : bool
+        see `decode(...) <pyjson5.decode_>`_
+
+    Raises
+    ------
+    Json5DecoderException
+        An exception occured while decoding.
+    TypeError
+        An argument had a wrong type.
+
+    Returns
+    -------
+    object
+        see `decode(...) <pyjson5.decode_>`_
+    '''
+    return decode_buffer(data, maxdepth, bool(some), 0)
+
+
 def decode_buffer(object obj, object maxdepth=None, object some=False,
                   object wordlength=None):
     '''
@@ -111,7 +143,7 @@
     some : bool
         see `decode(...) <pyjson5.decode_>`_
     wordlength : Optional[int]
-        Must be 1, 2, 4 to denote UCS1, USC2 or USC4 data.
+        Must be 0, 1, 2, 4 to denote UTF-8, UCS1, USC2 or USC4 data, resp.
         Surrogates are not supported. Decode the data to an ``str`` if need be.
         If ``None`` is supplied, then the buffer's ``itemsize`` is used.
 
@@ -574,7 +606,7 @@
 
 __all__ = (
     # DECODE
-    'decode', 'decode_latin1', 'decode_buffer', 'decode_callback', 'decode_io',
+    'decode', 'decode_utf8', 'decode_latin1', 'decode_buffer', 'decode_callback', 'decode_io',
     # ENCODE
     'encode', 'encode_bytes', 'encode_callback', 'encode_io', 'encode_noop', 'Options',
     # LEGACY
diff --git a/src/_imports.pyx b/src/_imports.pyx
index 139b356..34e8ca3 100644
--- a/src/_imports.pyx
+++ b/src/_imports.pyx
@@ -198,6 +198,11 @@
     boolean expect 'JSON5EncoderCpp_expect'(boolean actual, boolean expected)
 
 
+cdef extern from * nogil:
+    enum:
+        CYTHON_COMPILING_IN_PYPY
+
+
 cdef type datetime, date, time, Decimal, Mapping, IOBase
 cdef object saferepr
 
diff --git a/src/_reader_ucs.pyx b/src/_reader_ucs.pyx
index f94a5f2..bf187dc 100644
--- a/src/_reader_ucs.pyx
+++ b/src/_reader_ucs.pyx
@@ -19,9 +19,15 @@
     const Py_UCS4 *string
 
 
+cdef struct ReaderUTF8:
+    ReaderUCS base
+    const Py_UCS1 *string
+
+
 ctypedef ReaderUCS1 &ReaderUCS1Ref
 ctypedef ReaderUCS2 &ReaderUCS2Ref
 ctypedef ReaderUCS4 &ReaderUCS4Ref
+ctypedef ReaderUTF8 &ReaderUTF8Ref
 
 ctypedef Py_UCS1 *UCS1String
 ctypedef Py_UCS2 *UCS2String
@@ -31,6 +37,7 @@
     ReaderUCS1Ref
     ReaderUCS2Ref
     ReaderUCS4Ref
+    ReaderUTF8Ref
 
 ctypedef fused UCSString:
     UCS1String
@@ -50,3 +57,32 @@
     self.base.position += 1
 
     return cast_to_uint32(c)
+
+
+cdef inline uint32_t _reader_utf8_get(ReaderUCSRef self):
+    cdef uint32_t c0 = _reader_ucs_get(self)
+    cdef unsigned int n
+
+    if (c0 & 0b1_0000000) == 0b0_0000000:    # ASCII
+        return c0
+    elif (c0 & 0b11_000000) == 0b10_000000:  # broken continuation
+        return c0
+    elif (c0 & 0b111_00000) == 0b110_00000:  # 2 bytes
+        c0 = (c0 & 0b000_11111)
+        n = 1
+    elif (c0 & 0b1111_0000) == 0b1110_0000:  # 3 bytes
+        c0 = (c0 & 0b0000_1111)
+        n = 2
+    elif (c0 & 0b11111_000) == 0b11110_000:  # 4 bytes
+        c0 = (c0 & 0b00000_111)
+        n = 3
+    else:  # 5+ bytes, invalid
+        return c0
+
+    for n in range(n, 0, -1):
+        if not _reader_ucs_good(self):
+            return c0
+
+        c0 = (c0 << 6) | (_reader_ucs_get(self) & 0b00_111111)
+
+    return c0
diff --git a/src/_readers.pyx b/src/_readers.pyx
index fff1ba0..e0a910e 100644
--- a/src/_readers.pyx
+++ b/src/_readers.pyx
@@ -24,10 +24,14 @@
 
 
 cdef inline uint32_t _reader_get(ReaderRef self):
-    if ReaderRef in ReaderUCSRef:
-        return _reader_ucs_get(self)
+    cdef uint32_t c0
+    if ReaderRef is ReaderUTF8Ref:
+        c0 = _reader_utf8_get(self)
+    elif ReaderRef in ReaderUCSRef:
+        c0 = _reader_ucs_get(self)
     elif ReaderRef is ReaderCallbackRef:
-        return _reader_Callback_get(self)
+        c0 = _reader_Callback_get(self)
+    return c0
 
 
 cdef int32_t _reader_good(ReaderRef self) except -1: