I played around a bit with Io not too long ago, after someone recommended it to me over python (in general) for doing async network i/o. It seemed like a pretty solid language and had a decent library, too.
I don't have much to say about Ioke other than the fact that, personally, I doubt I would have chosen to re-write the language for some arguably minor changes. Of course, I'm sure the article doesn't get into all the differences.
It seems to be pretty much close to the speed of similar languages (python, ruby, etc). It loses a lot in the languages shootout[0], but that was back in January and its a very young language. It does seem to beat python in vector ops[1], so who knows what else it wins at.
Anyways, speed isn't the reason he is rewriting Io for the JVM - he wants the advantage of all the libraries already available, as well as being able to fix some small things he doesn't like.
I've been hanging around on the mailing list for a while, and it's clear to me that there has been very little work going on in Io in past months. (E.g.: There are obvious bugs in the Range addon that they've known about for as long as I've been using Io.) I'd be very doubtful that we've seen any performance increases, even small ones.
The vector benchmarks show off the sort of code you wouldn't really write with Io anyway, are three years old, against Python three versions ago.
You're right that Bini doesn't seem to care about performance here, and that's fine. But it's certainly a huge problem for Io, obvious even in small scripts, that would be reason enough to fork it.
I don't have much to say about Ioke other than the fact that, personally, I doubt I would have chosen to re-write the language for some arguably minor changes. Of course, I'm sure the article doesn't get into all the differences.