In roleplay context i find this release more consistent than the newer "refined" version...
The refined version seems very eager to jump into whatever you throw at it, like enthusiastically eager. This version remains skeptical to extreme suggestions that come out of the blue, which imo makes it a better fit to portray "people" believably. Perhaps the refined is a better fit for general assistant tasks, but for creative writing I'm sticking with this one for now.
I've tried out a bunch of abliterated gemma 3 versions, and this one is the most stable one both in terms of grammar (common issue with gemma 3 abliterations...), and in terms of general dialogue consistency\correct term usage\understanding subtext. The more popular (and somewhat older) ones that people recommend on forums didn't do it for me in those regards, while this one did, so thank you! Using Q5_K_M quants for comparisons.
Thank you for the feedback. The refined one was intended to be enthusiastically eager - to address some user's complaints that abliterated Gemma still tends to be a bit resistant. But I can totally understand how that eagerness to please may not fit everyone's style - I've had that same complaint about models I've roleplayed with in the past.