I switched to mollom for antispam on my blog one week ago, using the wp-mollom plugin. I wanted to give it a week so it'd get a fair shake, and figured anything had to do better than Akismet and SK2 were doing on my blog.
There was an initial warming up period for the first couple of days - I didn't realize this, but apparently my blog attracts a particular dialect of spam that is different from what had been seen by mollom before. After teaching it about these silly spammers (most appear to be based in eastern Europe, and use fragments of text from my own blog posts and comments to appear legit) mollom started to do pretty well.
The mollom service tracks stats on "ham" vs. "spam" and it is definitely doing better - unfortunately, the "ham" stats also appears to include comments that I had to moderate manually so the stats are a bit off kilter.
The initial green spike of "ham" on August 6 was actually almost entirely made up of comments that got through mollom and that I had to manually moderate as spam. My current mollom stats indicate:
So far, Mollom has blocked or moderated 3803 messages on your website of which you moderated 4.92% yourself.
Over the entire week, there were only 35 legitimate comments posted to the blog. My current stats say there have been 3803 messages yanked as spam (either directly by mollom, or manually). That means over 99% of comments that were attempted to be posted on my blog were spam. mollom's stats say I've manually moderated 4.92% of the spam comments - that's 187 comments manually spanked. 26 per day. But it's definitely getting better.
There are still occasional spam attacks (most notably by an apparent ring of eastern European spamroaches), but by and large, mollom is already outperforming both Akismet and SK2 on my blog. I'll be sticking with mollom for awhile - hopefully the trend continues and it keeps getting smarter and more effective at dealing with the particular variants of spam that get sent my way.
The wp-mollom plugin could use some minor refinement - much of the spam I get reuses bits of text from my own blog posts and previous comments on that post to appear legitimate. It would be much easier to detect this if a link to the post was provided on the mollommanage page.
It would also be much more effective if the referring web page and user agent were indicated on a comment so I could see at a glance if it came from a spambot, or was sent by the various pagerank backlink checking utilities - both obvious signs of spamass activity.
I'd posted some previous feedback on the mollom forums, and it looks like the suggestions have already been implemented. That's fantastic.