---
license: apache-2.0
base_model: Qwen/Qwen2.5-3B-Instruct
pipeline_tag: text-generation
library_name: rkllm
tags:
- rkllm
- rockchip
- rk3576
- rk3588
- qwen
- qwen2.5
---

# Qwen2.5-3B-Instruct-RKLLM

RKLLM-converted Qwen2.5-3B-Instruct language-model artifacts for Rockchip RK3576 and RK3588 NPUs.

These hardware-specific `.rkllm` files require a compatible Rockchip RKLLM runtime. They are not Transformers checkpoints and cannot be loaded directly with Transformers, llama.cpp, or Ollama.

## Base model

- Upstream model: [Qwen/Qwen2.5-3B-Instruct](https://huggingface.co/Qwen/Qwen2.5-3B-Instruct)
- License: Qwen Research License
- Model type: LLM (text only)

Review the upstream license and usage restrictions before use or redistribution.

## Conversion and variants

### Toolkit version

**RKLLM Toolkit: v1.2.3**

Use a file built for the exact target SoC.

| Target | Quantization | File | SHA256 |
| --- | --- | --- | --- |
| RK3576 | W4A16 | [Qwen2.5-3B-Instruct_RK3576_w4a16.rkllm](RK3576/Qwen2.5-3B-Instruct_RK3576_w4a16.rkllm) | `5f2480e10a794848c8d4a5a21a61c015d96570f2a58e4543897b694b59576908` |
| RK3576 | W8A8 | [Qwen2.5-3B-Instruct_RK3576_w8a8.rkllm](RK3576/Qwen2.5-3B-Instruct_RK3576_w8a8.rkllm) | `5a14ed85f65d3c2890ef8e2b4ab9f094bf9b0992ddebd6f2b9ad8d3b539897b9` |
| RK3588 | W8A8 | [Qwen2.5-3B-Instruct_RK3588_w8a8.rkllm](RK3588/Qwen2.5-3B-Instruct_RK3588_w8a8.rkllm) | `054a4ac54ea7d483ac17431df5286eb6b2a81d351fbb44b55dc4491f71a7ea46` |

The repository also includes `Qwen2.5-3B-Instruct_data_quant.json`, used as calibration data during conversion.

## Usage

```bash
hf download HanzoHuang/Qwen2.5-3B-Instruct-RKLLM \
  RK3576/Qwen2.5-3B-Instruct_RK3576_w4a16.rkllm \
  --local-dir Qwen2.5-3B-Instruct-RKLLM
```

Use the Qwen2.5 Instruct chat template with the RKLLM runtime. For Docker deployment, see [Hanzo-Huang/rkllm-docker](https://github.com/Hanzo-Huang/rkllm-docker).

## Limitations

These are target-specific converted artifacts. Validate quality and runtime compatibility on your Rockchip device.

## Acknowledgements

Thanks to the Qwen Team, Rockchip, and the RKLLM community.