MergekitCloud/mergekit-102
MergekitCloud/mergekit-102 is an 8 billion parameter language model created by MergekitCloud using the linear merge method. This model combines Qwen3-8B, ertghiu256/qwen3-8b-code-reasoning, and deepseek-ai/DeepSeek-R1-0528-Qwen3-8B, leveraging their strengths. With a 32768 token context length, it is designed for enhanced performance across general language tasks and specialized code reasoning applications.
Loading preview...
Overview
MergekitCloud/mergekit-102 is an 8 billion parameter language model developed by MergekitCloud. It was constructed using the linear merge method from the mergekit framework, combining the strengths of several pre-trained models. This approach aims to create a robust model by integrating diverse capabilities from its constituent parts.
Key Capabilities
- Enhanced General Language Understanding: Incorporates the base capabilities of Qwen3-8B for broad language tasks.
- Specialized Code Reasoning: Benefits from the inclusion of
ertghiu256/qwen3-8b-code-reasoning, suggesting improved performance in code-related tasks and logical reasoning. - Integrated Strengths: Leverages contributions from
deepseek-ai/DeepSeek-R1-0528-Qwen3-8Bto further refine its overall performance.
Merge Details
The model was created by merging three distinct models with specific weighting parameters:
Qwen/Qwen3-8B(weight: 0.34)ertghiu256/qwen3-8b-code-reasoning(weight: 0.55)deepseek-ai/DeepSeek-R1-0528-Qwen3-8B(weight: 0.33)
The merge process normalized these weights and usedbfloat16for the data type, withQwen/Qwen3-8Bserving as the tokenizer source. The model supports an auto chat template.
Good For
This model is suitable for applications requiring a balance of general language understanding and specific strengths in areas like code reasoning, making it a versatile choice for developers looking for a merged 8B parameter model.