What the manual and the first commenter are trying to say is that mb_strcut uses byte offsets, as opposed to mb_substr which uses character offsets.
Both mb_strcut and mb_substr appear to treat negative and out-of-range offsets and lengths in the basically the same way as substr. An exception is that if start is too large, an empty string will be returned rather than FALSE. Testing indicates that mb_strcut first works out start and end byte offsets, then moves each offset left to the nearest character boundary.
(PHP 4 >= 4.0.6, PHP 5)
mb_strcut — Get part of string
Descriptionstring mb_strcut ( string $str, int $start [, int $length [, string $encoding]] )
mb_strcut() returns the portion of str specified by the start and length parameters.
mb_strcut() performs equivalent operation as mb_substr() with different method. If start position is multi-byte character's second byte or larger, it starts from first byte of multi-byte character.
It subtracts string from str that is shorter than length AND character that is not part of multi-byte string or not being middle of shift sequence.
encoding is character encoding. If it is not set, internal character encoding is used.
diffrence between mb_substr and mb_substr
mb_strcut('I_ROHA', 1, 2) returns 'I_'. Treated as byte stream.
mb_substr('I_ROHA', 1, 2) returns 'ROHA' Treated as character stream.
# 'I_' 'RO' 'HA' means multi-byte character