That's right: the UTF-8 encoding of the greek symbols are two bytes long. The first one will be (in binary) 110xxxxx and the second one 10xxxxxx, where the x's are the actual code point value up to 11 bits. In octal, that means a leading 3 and a leading 2, respectivly.

By "as expected", you mean the same results I got, not the previosly expected results as defined in the docs, right?

The funny chars in the error message are due to the Console window using a different code page. It's using a DOS-compatible OEM code page, probably 437. You can change that via "MODE CON CP SELECT=1251" to match the GUI, but that won't help here since there is no UTF-8 setting. Redirect it to a file and view with a UTF-8 editor.