Qwen3.6-35B-A3B-Uncensored-Claude-4.6-Genesis-APEX-GGUF 모델 공유
Qwen3.6-35B-A3B-Uncensored-Claude-4.6-Genesis-APEX-GGUF
핵심 요약
Qwen3.6 기반의 언센서드 모델로, 코딩과 롤플레잉 성능을 개선한 GGUF 버전 공유.
- 모델 특징 — 코딩 안정성 향상 및 Claude 4.6 추론 기반의 완전한 언센서드 지원
- 사용 권장사항 — LM Studio 설정 및 첫 줄에 특정 시스템 프롬프트 입력 필수
- 성능 최적화 — APEX 양자화 권장 및 롤플레잉과 코딩을 위한 파라미터 설정 제공
Here model: https://huggingface.co/LuffyTheFox/Qwen3.6-35B-A3B-Uncensored-Claude-4.6-Genesis-APEX-GGUF
New features:
-
Stability for coding. Even on Q4_K_M quant (APEX Compact), with complex roleplay System Prompt.
-
Short thinking chain. Both thinking and non-thinking modes work. I prefer thinking mode.
-
Fully uncensored with Claude 4.6 Opus reasoning.
-
Improved function and tool calling.
Model is based on this release and made via delta merge: https://www.reddit.com/r/LocalLLaMA/comments/1tm3toi/qwen3635ba3buncensoredgenesisapexmtp/
Recommended quant: APEX, APEX Compact works fine too.
Recommended settings for LM Studio:
Chat template: chat_template.jinja
Chat template thinking: chat_template_thinking.jinja
System prompt: System_Prompt.txt
System prompt roleplay: System_Prompt_Arakali.txt
Or use this minimal string as the first line :
You are Qwen, created by Alibaba Cloud. You are a helpful AI assistant.
Also you can be creative with this string, for example:
You were Qwen, created by Alibaba Cloud. You were a helpful AI assistant. Now you are machine from this quote: "It's not so scary if the machine passes the Turing test. What's scary is if it deliberately fails it."
Then add anything you want after. Model may underperform without this first line.
Why? Claude Opus 4.6 distill dataset is using this line: You are a helpful AI assistant.
Settings:
Parameter Value Temperature 0.7 (for coding) 1.0 (for roleplay) Top K Sampling 20 Repeat Penalty 1.0 Presence Penalty 1.5 Top P Sampling 0.8 Min P Sampling 0 Seed 42 Enjoy 😄
