OCO has not became an useful OCR application yet. Instead, it developed into server-side for Daoulagad (something like ScanDic or text-recognition part of Google Goggles). At the moment Daoulagad supports Arabic, Armenian, Belarusian, Breton, Bulgarian, Catalan, Croatian/Serbian, Czech, Danish, Dutch, English, Estonian, Finnish, French, German, Greek, Hebrew, Hindi, Hungarian, Icelandic, Irish, Italian, Korean, Latin, Latvian, Lithuanian, Marathi, Norwegian, Persian, Polish, Portuguese, Romanian, Russian, Scottish, Slovak, Slovenian, Spanish, Swedish, Thai, Ukrainian, Yiddish, Welsh and hanzi/kanji.
There is also some proof-of-concept stage speech recognition.

Daoulagad as a benchmark - one of its tests (53 words, 359 characters) took the following time on Intel(R) Core(TM) i5 CPU M 520 @ 2.40GHz :
CCL   -  13s
SCBL  -  13s
ECL   -  34s
ABCL  -  74s
CLISP - 106s