First off, I am aware of Mercurial: Problem with non-ascii letters in filenames between Windows and Linux and What DVCS support Unicode filenames?. But they are old, and do
No, Mercurial still doesn't support transcoding of filenames. That is, it will checkin and checkout filenames as binary strings and you will experience problems if you need to move files between systems with incompatible filename encodings.
If everybody in your team is using Windows with the same code page, then you can use non-ASCII characters just fine. They'll be stored inside the repository in whatever code page you use, but everybody will be able to use the files.
If you need to move files between platforms, then stick to ASCII filenames for now.
There is a Windows UTF-8 plan in the works, and FUJIWARA Katsunori has been working on it, but it's not yet (September 2012) ready.