metadata

base_model:
  - burgasdotpro/Sebushka-llama-3.1-8B
  - Vikhrmodels/Vikhr-Llama3.1-8B-Instruct-R-21-09-24
library_name: transformers
tags:
  - mergekit
  - merge
  - roleplay
  - assistant
language:
  - ru

merge

This is a merge of pre-trained language models

This merge is fully focused on merging good, all purpose RU 8B model, for people with low VRAM/RAM.

After many tries, and even more testing, i could say that this model performs in russian as good as llama 8B can.

Model could be used in roleplay, as assistant, it even tries to follow instructions.

Of course, this is 8B, so don't expect much from it.