tepirale/Qwen3.5-27B-ThinkSonic-Coder-27B-dare_ties-MTP
tepirale/Qwen3.5-27B-ThinkSonic-Coder-27B-dare_ties-MTP is a 27 billion parameter language model merge based on Qwen/Qwen3.5-27B, created using the DARE TIES method. This model integrates components from ThinkingCap-Qwen3.6-27B, Qwopus3.6-27B-Coder, and Qwen3.5-27B-Claude-4.6-Opus-Reasoning-Distilled. It is designed to enhance reasoning and coding capabilities, leveraging the Qwen3.5 MTP model architecture with a 32768 token context length.
Loading preview...
Model Overview
This model, tepirale/Qwen3.5-27B-ThinkSonic-Coder-27B-dare_ties-MTP, is a 27 billion parameter language model merge built upon the Qwen/Qwen3.5-27B base. It was created using the DARE TIES merge method, a technique for combining pre-trained language models. The merge specifically incorporates the MTP (Multi-Task Pre-training) model from Qwen3.5, with its layers left unchanged to test the performance of the new DARE TIES algorithm.
Key Capabilities & Features
- Architecture: Based on the Qwen3.5-27B model, known for its robust performance.
- Merge Method: Utilizes the DARE TIES algorithm, which is designed for efficient and effective model merging.
- Component Models: Integrates specialized models:
bottlecapai/ThinkingCap-Qwen3.6-27BJackrong/Qwopus3.6-27B-CoderJackrong/Qwen3.5-27B-Claude-4.6-Opus-Reasoning-Distilled
- Focus: The inclusion of "Coder" and "Reasoning-Distilled" components suggests an emphasis on coding proficiency and advanced reasoning tasks.
- Context Length: Supports a context length of 32768 tokens.
Intended Use Cases
This model is particularly well-suited for applications requiring:
- Code Generation and Understanding: Leveraging the
Qwopus3.6-27B-Codercomponent. - Complex Reasoning Tasks: Benefiting from the
ThinkingCapandClaude-4.6-Opus-Reasoning-Distilledinfluences. - Research into Model Merging: Serves as a testbed for the DARE TIES algorithm's performance with the Qwen3.5 MTP architecture.
Initial evaluations indicate good performance for this merged model.