Much better than llama 70b
Title. I think that Llama 3.3 is definitely worse than 3.1 in RP.
This model is much smarter than all Llama versions and don't spit "happy" slop every message like llama does.
I sometimes think that stheno is smarter than eureale in rp.
I disagree. Llama 3.3 might have better real talk understanding. but the karens at Meta trained Llama 3.3 off a cliff, being over zealous with safety. Llama is Taylor perfect for a 7 year old.
Steheno is a pimp God.
Kunou 70b is a basic bitch.
Im only able to run Kunou on int4 I don't have enough vram to run on int8 or 16, does anyone know if that makes much difference on the Llama pixie dust?
Title. I think that Llama 3.3 is definitely worse than 3.1 in RP.
This model is much smarter than all Llama versions and don't spit "happy" slop every message like llama does.
I sometimes think that stheno is smarter than eureale in rp.
I fully agree, this model is much better than Llama 3.3. For my personal taste, Llama speaks too obvious and corny compared to Qwen, who pays attention to smaller and more realistic details in her dialogue. I noticed the same thing with Eva Qwen and Eva Llama, I liked Qwen much more.