Qwen3-4B-Deconstruct-V2-GGUF : GGUF

This model was finetuned and converted to GGUF format using Unsloth.

Example usage:

• For text only LLMs: ./llama.cpp/llama-cli -hf RinKana/Qwen3-4B-Deconstruct-V2-GGUF --jinja
• For multimodal models: ./llama.cpp/llama-mtmd-cli -hf RinKana/Qwen3-4B-Deconstruct-V2-GGUF --jinja

Reasoning Tag Format

The model uses a specific Tetralemma/Deconstructionist XML structure:

<reason> initial explanation </reason>
<exception> counterfactual explanation </exception>
<tension> divergence explanation </tension>
<categorization> rule & exception segregation </categorization>
<deconstruction> rule & exception refutation </deconstruction>
<conclusion> recap of entire sequence </conclusion>

Available Model files:

• qwen3-4b.Q4_K_M.gguf

Ollama

An Ollama Modelfile is included for easy deployment. This was trained 2x faster with Unsloth