---
base_model:
- Qwen/Qwen3.5-4B
tags:
- text-generation-inference
- transformers
- unsloth
- qwen3_5
license: apache-2.0
language:
- en
- vi
- zh
datasets:
- beyoru/Aesir-Character-CoT-roleplay
- beyoru/Aisaka-LN-v4
library_name: transformers
---

# Rhapsody-4B

**Rhapsody-4B** is a fine-tuned language model built on **Qwen3.5-4B**. It is trained on a mixture of seven instruction datasets, with sampling prioritized in the following order: roleplay, agent, coding, science, finance, and logical reasoning. The model is designed to improve instruction following while preserving the general capabilities of the base model.

<p align="center">
  <img src="beyoru.png" width="512">
</p>

## Citation

If you use **Rhapsody-4B** in your work, please cite:

```bibtex
@misc{rhapsody4b2026,
  title        = {Rhapsody-4B},
  author       = {Beyoru},
  year         = {2026},
  howpublished = {\url{https://huggingface.co/beyoru/Rhapsody-4B}}
}
```

## License

Apache-2.0