README.md

metadata

title: Qwen3-4B Claude Reasoning
emoji: 🧠
colorFrom: indigo
colorTo: pink
sdk: gradio
pinned: true

Qwen3-4B Claude Sonnet Reasoning Distill (GGUF Q8_0)

This model was trained on a Claude Sonnet 4 (non-reasoning) dataset and a Claude Sonnet 3.7 (reasoning) dataset.

• 🧬 Datasets:

  • Liontix/claude-sonnet-4-100x
  • reedmayhew/claude-3.7-sonnet-reasoning

• 🏗 Base Model:

  • unsloth/Qwen3-4B-unsloth-bnb-4bit

If you want to fine-tune this model:

• Start from: Liontix/Qwen3-4B-Claude-Sonnet-4-Reasoning-Distill-Safetensor
• Change dataset as needed in your training script or notebook

Prompt format uses Claude-style <|im_start|> / <|im_end|> markers with role tags.