File size: 1,971 Bytes
705d9ea 3d9ed64 705d9ea 3d9ed64 4603140 3d9ed64 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 | ---
license: apache-2.0
base_model: openbmb/MiniCPM3-4B
pipeline_tag: text-generation
library_name: rkllm
tags:
- rkllm
- rockchip
- rk3576
- rk3588
- minicpm3
---
# MiniCPM3-4B-RKLLM
RKLLM-converted MiniCPM3-4B language-model artifacts for Rockchip RK3576 and RK3588 NPUs.
These are hardware-specific `.rkllm` binaries, not Transformers checkpoints. They require a compatible Rockchip RKLLM runtime and cannot be loaded directly with Transformers, llama.cpp, or Ollama.
## Base model
- Upstream model: [openbmb/MiniCPM3-4B](https://huggingface.co/openbmb/MiniCPM3-4B)
- License: Apache-2.0
- Model type: LLM (text only)
## Conversion and variants
### Toolkit version
**RKLLM Toolkit: v1.3.0**
Use a file built for the exact target SoC.
| Target | Quantization | File | SHA256 |
| --- | --- | --- | --- |
| RK3576 | W4A16 (g128) | [MiniCPM3-4B_RK3576_w4a16_g128.rkllm](RK3576/MiniCPM3-4B_RK3576_w4a16_g128.rkllm) | `ec05211d17c4bb05a685588ff335608233b59722f6ecc296976d61dbc88c4128` |
| RK3576 | W8A8 | [MiniCPM3-4B_RK3576_w8a8.rkllm](RK3576/MiniCPM3-4B_RK3576_w8a8.rkllm) | `3795c6a22e5f12de0c4243fc187da55f516b530de3125734874d340856160aec` |
| RK3588 | W8A8 | [MiniCPM3-4B_RK3588_w8a8.rkllm](RK3588/MiniCPM3-4B_RK3588_w8a8.rkllm) | `3815a59593df7aebda90f86844caf2fd625d9d0e996e35aa31fd74790d10b130` |
The repository also includes `MiniCPM3-4B_data_quant.json`, used as calibration data during conversion.
## Usage
```bash
hf download HanzoHuang/MiniCPM3-4B-RKLLM \
RK3576/MiniCPM3-4B_RK3576_w4a16_g128.rkllm \
--local-dir MiniCPM3-4B-RKLLM
```
Run the downloaded file with the RKLLM runtime and the upstream MiniCPM3 chat template. For Docker deployment, see [Hanzo-Huang/rkllm-docker](https://github.com/Hanzo-Huang/rkllm-docker).
## Limitations
These artifacts are target-specific conversions. Validate quality, memory use, and runtime compatibility on your own hardware.
## Acknowledgements
Thanks to OpenBMB, Rockchip, and the RKLLM community.
|