---
license: apache-2.0
datasets:
- nvidia/OpenCodeReasoning
- vicgalle/creative-rubrics-gpt-4.5-o3-R1
base_model:
- unsloth/Qwen3-4B
tags:
- unsloth
- trl
- sft
- code
pipeline_tag: text-generation
---
# Qwen 3 Code Reasoning 

A small Qwen 3  4 billion parameter model trained on [nvidia/OpenCodeReasoning](https://huggingface.co/datasets/nvidia/OpenCodeReasoning) for coding tasks.
For Coding, it is recommended to be in thinking mode.  

## Strengths 
- Code generation
- Logical question answering

## Drawbacks 
- Heavy overthinking
- Context overflow

## Recommended Usage: 
- [vllm](https://github.com/vllm-project/vllm)
- [transformers](https://github.com/QwenLM/Qwen3?tab=readme-ov-file#run-qwen3)

### GGUF VERSION
[gguf](https://huggingface.co/ertghiu256/qwen3-4b-code-reasoning-gguf)