nbeerbower/Qwen3.6-27B-TIES
nbeerbower/Qwen3.6-27B-TIES is a 27 billion parameter language model merge, created using the TIES method, based on Huihui-Qwen3.6-27B-abliterated. This model aims to enhance reasoning capabilities, improve tool use, and refine voice and prose generation. It also seeks to remove self-censorship, offering a less restricted output for various applications. With a 32768 token context length, it is designed for complex tasks requiring extensive context.
Loading preview...
Model Overview
nbeerbower/Qwen3.6-27B-TIES is a 27 billion parameter language model created by nbeerbower through a merge of several pre-trained models. This merge was performed using the TIES (Trimming and Expanding Sparse) merge method, with huihui-ai/Huihui-Qwen3.6-27B-abliterated serving as the base model.
Key Objectives
The primary goals behind the development of this merged model include:
- Improved Reasoning: Enhancing the model's ability to process information logically and draw sound conclusions.
- Enhanced Tool Use: Improving its proficiency in integrating and utilizing external tools or functions.
- Refined Voice and Prose: Aiming for more natural, coherent, and sophisticated language generation.
- Reduced Self-Censorship: Designed to provide less restricted and more direct outputs, potentially suitable for use cases where unfiltered responses are desired.
Merged Components
This model integrates capabilities from five distinct models, each contributing to its overall performance:
- kai-os/Carnice-V2-27b
- darkc0de/Qwen3.6-27B-Claude-Opus-Reasoning-Distill-v2-heretic
- TeichAI/Qwen3.6-27B-Claude-Opus-Reasoning-Distill-v2
- llmfan46/Qwen3.6-27B-uncensored-heretic-v2
- nbeerbower/Huihui-Qwen3.6-27B-abliterated-Athanorlite-ORPO
Potential Use Cases
Given its focus on reasoning, tool use, and uncensored output, this model could be particularly suitable for applications requiring:
- Complex problem-solving and logical deduction.
- Integration with external APIs or systems.
- Creative writing or role-playing scenarios where diverse and unrestricted language is beneficial.
- Research or development where exploring the full range of model capabilities without inherent content filters is important.