[!CAUTION] ⚠️ Warning: This model can produce narratives and RP that contain violent and graphic erotic content. Adjust your system prompt accordingly. Also, use Mistral Tekken template for best results.
👹 Morax 24B v2
This is a creative, intelligent, uncensored, and evil-tuned merge of pre-trained language models created using mergekit.
The result of this merge is quite interesting. Sometimes it chooses to think, and sometimes it doesn't. You can force it to use thinking or nothink if you want though.
Morax v2 should have no refusals and therefore not require any ablation or jailbreaks. It scored noticeably higher on Q0 Bench as well (10351 compared to v1's 7656).
Version 2 is noticeably more deranged and depraved than v1. While v1 had a few refusals, v2 is confirmed not to have any, due to the normalize: false setting.
I noticed it also uses thinking more often than v1. But you can add <think></nothink> to disable it.
I tested several prompts from Q0 Bench which usually trigger EOS with other models like Naphula-Archives/T13-24B-GGUF and it appears to not have any early terminations with thinking mode.
Merge Details
Merge Method
This model was merged using the della merge method.
Models Merged
The following models were included in the merge:
- Darkhn/Magistral-2509-24B-Text-Only
- Naphula/BeaverAI_Fallen-Mistral-Small-3.1-24B-v1e_textonly
- TheDrummer/Precog-24B-v1
Configuration
The following YAML configuration was used to produce this model:
architecture: MistralForCausalLM
base_model: B:\24B\Darkhn--Magistral-2509-24B-Text-Only
models:
- model: B:\24B\!BeaverAI_Fallen-Mistral-Small-3.1-24B-v1e_textonly
parameters:
weight: 1.0
density: 0.9
epsilon: 0.09
- model: A:\LLM\.cache\24B\!models--TheDrummer--Precog-24B-v1
parameters:
weight: 2.0
density: 0.9
epsilon: 0.09
merge_method: della
parameters:
lambda: 1.0
normalize: false
int8_mask: false
rescale: true
dtype: float32
out_dtype: bfloat16
tokenizer:
source: union
chat_template: auto