Liontix/Qwen3-4B-Claude-Sonnet-4-Reasoning-Distill-Safetensor is a 4 billion parameter language model based on the Qwen3 architecture, fine-tuned by Liontix. It was distilled from a combination of Claude Sonnet 4 (non-reasoning) and Claude Sonnet 3.7 (reasoning) datasets, making it specialized for tasks requiring reasoning capabilities. This model leverages a 40960 token context length and is designed for applications benefiting from Claude-style instruction following and distilled reasoning patterns.
Loading preview...
Model Overview
Liontix/Qwen3-4B-Claude-Sonnet-4-Reasoning-Distill-Safetensor is a 4 billion parameter language model built upon the unsloth/Qwen3-4B-unsloth-bnb-4bit base model. This model is specifically designed to distill reasoning capabilities from advanced large language models.
Key Characteristics
- Distilled Reasoning: The model was fine-tuned using a unique combination of datasets:
Liontix/claude-sonnet-4-100x(derived from Claude Sonnet 4, focusing on non-reasoning aspects)reedmayhew/claude-3.7-sonnet-reasoning(derived from Claude Sonnet 3.7, specifically targeting reasoning capabilities)
- Claude-Style Prompting: It utilizes Claude-style prompt formatting with
<|im_start|>and<|im_end|>markers, including role tags, for instruction following. - Base Architecture: Leverages the Qwen3 architecture, known for its efficiency and performance.
Use Cases
This model is particularly well-suited for applications where:
- Reasoning tasks are critical, benefiting from the distillation of Claude Sonnet 3.7's reasoning patterns.
- Claude-style instruction following is preferred or required for consistent interaction.
- A 4 billion parameter model with a substantial context length (40960 tokens) offers a balance between performance and computational efficiency.