Dampfinchen/Ultra-Instruct-12B

Text Generation • Updated Aug 22 • 10 • 4

Note Performs extremely well! It seems to be a mix of Instruct and RP focused data leading to some really amazing results. XTC sampling was also recently added to KoboldCpp and it makes the model even better. Strongly recommend this model over any of the merges I've done.

starble-dev/Hollow-Tail-V1-12B

Text Generation • Updated Aug 19 • 11 • 3

Note A merge of two RP models plus an intelligence model. Performs quite well and even seems to remember events previously mentioned. Not sure if this is the norm but I like it. Definitely not perfect given it has quirks with point of view changing but it's still quite solid.