extremely good model
this model has exceeded my expectations and performs better than the recent L3 model breakthroughs. a diamond in the rough so to speak, i didn't expect to find a small sized model that wasn't L3 that was capable of performing as well as QuartetAnemoi.
this model performs extremely well with large context character cards though it does have its quirks, it does need corrected on occasion but you can really tell it does it's best to follow rules that are neatly laid out.
9 out of 10 stars.
settings used:
temperature 1.2
top p 1
top k 30
repeat penalty 1.05
repeat penalty tokens at 250
prompt template Model default
UI used faraday/backyardAI
Been using this on a IRC chatbot and so far it is only model that can handle interacting with many different people without getting confused.