---
license: cc-by-nc-4.0
---
~~I don't think GGUF version can use. Haven't test yet. But expect this is very good model.~~

Inixion-v2, My best MoErge model for now, have a capable of RP and general task. It can:

- Understand well the character card
- Smart model, so have very well logic
- You can play multi character if you like
- Could use RoPe scale to 16k context length, but hey, maybe it unstable.

Some still not good:

- 8B model base, so although it 13B(24B?) parameter now with MoE merge, it still lack of knowledge and sometime cause trouble
- Unexpected bug maybe occurs
- Big model, so need at least 8GB vram to run, better use 12GB Vram to load GGUF version
- It not too violence and toxic (maybe I need place this to good feature? Because not all everyone want toxic model w(ﾟДﾟ)w )

# GGUF?
https://huggingface.co./Alsebay/Inixion-2x8B-v2-GGUF

My own GGUF quantz, please use fixed version.

you can find other GGUF version in my friend https://huggingface.co./mradermacher

Here is imatrix GGUF version https://huggingface.co./mradermacher/Inixion-2x8B-v2-i1-GGUF

# Forgot to tell the recipe:
 
base_model:  MaziyarPanahi/Llama-3-8B-Instruct-v0.9

source_model:  NeverSleep/Llama-3-Lumimaid-8B-v0.1

source_model: Sao10K/L3-8B-Stheno-v3.2

All is very smart model, so you could enjoy it