Open-weight model
RK3576-LLM-Qwen3-4B-Instruct-2507
by RKNNAI RKNNAI/RK3576-LLM-Qwen3-4B-Instruct-2507
RK3576-LLM-Qwen3-4B-Instruct-2507 is an open-weight model from RKNNAI, released under Apache License 2.0. Its published files total 27.0 GB.
This repository provides RKLLM deployment configurations of Qwen3-4B-Instruct-2507 for RK3576. From the directory where you ran the download command, enter the configuration directory and verify: All entries must report OK before deployment.
Model Card
By RKNNAI, published under apache-2.0, revision 9c7142efe5b3.
This repository provides RKLLM deployment configurations of Qwen3-4B-Instruct-2507 for RK3576. From the directory where you ran the download command, enter the configuration directory and verify: All entries must report OK before deployment. - Supported chips are listed per configuration above. - Use the matching files from the same configuration.
Read RKNNAI's full model card
Qwen3-4B-Instruct-2507
English | 简体中文
1. Model Introduction
This repository provides RKLLM deployment configurations of Qwen3-4B-Instruct-2507 for RK3576.
| Field | Value |
|---|---|
| Model ID | RKNNAI/RK3576-LLM-Qwen3-4B-Instruct-2507 |
| Display Name | RK3576-LLM-Qwen3-4B-Instruct-2507 |
| Source Model | https://huggingface.co/Qwen/Qwen3-4B-Instruct-2507 |
| Model Type | LLM |
Available Models
| Configuration | Supported Chips | RKLLM Runtime Version | Quantization | NPU Cores | Context |
|---|---|---|---|---|---|
| Qwen3-4B-Instruct-2507-w4a16-2-1024 | RK3576 | v1.2.4 |
w4a16 | 2 | 1024 (1k) |
| Qwen3-4B-Instruct-2507-w4a16-2-16384 | RK3576 | v1.2.4 |
w4a16 | 2 | 16384 (16k) |
| Qwen3-4B-Instruct-2507-w4a16-2-2048 | RK3576 | v1.2.4 |
w4a16 | 2 | 2048 (2k) |
| Qwen3-4B-Instruct-2507-w4a16-2-4096 | RK3576 | v1.2.4 |
w4a16 | 2 | 4096 (4k) |
| Qwen3-4B-Instruct-2507-w8a8-2-1024 | RK3576 | v1.2.4 |
w8a8 | 2 | 1024 (1k) |
| Qwen3-4B-Instruct-2507-w8a8-2-2048 | RK3576 | v1.2.4 |
w8a8 | 2 | 2048 (2k) |
| Qwen3-4B-Instruct-2507-w8a8-2-4096 | RK3576 | v1.2.4 |
w8a8 | 2 | 4096 (4k) |
2. Files
| Path | Description |
|---|---|
README.md |
English documentation |
README_CN.md |
Chinese documentation |
LICENSE |
License |
Qwen3-4B-Instruct-2507-w4a16-2-1024/ |
Deployment configuration |
Qwen3-4B-Instruct-2507-w4a16-2-16384/ |
Deployment configuration |
Qwen3-4B-Instruct-2507-w4a16-2-2048/ |
Deployment configuration |
Qwen3-4B-Instruct-2507-w4a16-2-4096/ |
Deployment configuration |
Qwen3-4B-Instruct-2507-w8a8-2-1024/ |
Deployment configuration |
Qwen3-4B-Instruct-2507-w8a8-2-2048/ |
Deployment configuration |
Qwen3-4B-Instruct-2507-w8a8-2-4096/ |
Deployment configuration |
3. Download
ModelScope
Full repository:
modelscope download --model RKNNAI/RK3576-LLM-Qwen3-4B-Instruct-2507 --revision v1.2.4 --local_dir ./RK3576-LLM-Qwen3-4B-Instruct-2507
Selected configuration and root documents:
from modelscope import snapshot_download
snapshot_download(
"RKNNAI/RK3576-LLM-Qwen3-4B-Instruct-2507",
revision="v1.2.4",
allow_patterns=["README.md", "README_CN.md", "LICENSE", "Qwen3-4B-Instruct-2507-w4a16-2-1024/**"],
local_dir="./RK3576-LLM-Qwen3-4B-Instruct-2507",
)
Hugging Face
Full repository:
hf download RKNNAI/RK3576-LLM-Qwen3-4B-Instruct-2507 --revision v1.2.4 --local-dir ./RK3576-LLM-Qwen3-4B-Instruct-2507
Selected configuration and root documents:
hf download RKNNAI/RK3576-LLM-Qwen3-4B-Instruct-2507 --revision v1.2.4 --local-dir ./RK3576-LLM-Qwen3-4B-Instruct-2507 --include README.md README_CN.md LICENSE 'Qwen3-4B-Instruct-2507-w4a16-2-1024/**'
4. SHA-256 Verification
From the directory where you ran the download command, enter the configuration directory and verify:
cd ./RK3576-LLM-Qwen3-4B-Instruct-2507/Qwen3-4B-Instruct-2507-w4a16-2-1024
sha256sum -c SHA256SUMS
All entries must report OK before deployment.
5. Compatibility and Limitations
- Supported chips are listed per configuration above.
- Use the matching files from the same configuration.
6. Copyright and License
Source model: Qwen3-4B-Instruct-2507. License: apache-2.0; see LICENSE.
Identity and Version
- Repository
- RKNNAI/RK3576-LLM-Qwen3-4B-Instruct-2507
- Publisher
- RKNNAI
- Task
- Not stated by the source
- Modality
- Other
- Library
- Not stated by the source
- Parameters
- Not stated by the source
- Languages
- Not stated by the source
- Revision
- 9c7142efe5b343331981ae5a0ba0ae20a67cd3ab
- First published
- 2026-10-02
- Last updated
- 2026-10-02
Files and Weights
39 files, 27.0 GB in total.
Every file
| File | Type | Size | SHA-256 |
|---|---|---|---|
| Qwen3-4B-Instruct-2507-w4a16-2-1024/config.json | Configuration | 935 B | — |
| Qwen3-4B-Instruct-2507-w4a16-2-16384/config.json | Configuration | 936 B | — |
| Qwen3-4B-Instruct-2507-w4a16-2-2048/config.json | Configuration | 935 B | — |
| Qwen3-4B-Instruct-2507-w4a16-2-4096/config.json | Configuration | 935 B | — |
| Qwen3-4B-Instruct-2507-w8a8-2-1024/config.json | Configuration | 934 B | — |
| Qwen3-4B-Instruct-2507-w8a8-2-2048/config.json | Configuration | 934 B | — |
| Qwen3-4B-Instruct-2507-w8a8-2-4096/config.json | Configuration | 934 B | — |
| LICENSE | Documentation | 11.3 KB | — |
| Qwen3-4B-Instruct-2507-w4a16-2-1024/README.md | Documentation | 2.0 KB | — |
| Qwen3-4B-Instruct-2507-w4a16-2-1024/README_CN.md | Documentation | 1.9 KB | — |
| Qwen3-4B-Instruct-2507-w4a16-2-16384/README.md | Documentation | 2.0 KB | — |
| Qwen3-4B-Instruct-2507-w4a16-2-16384/README_CN.md | Documentation | 1.9 KB | — |
| Qwen3-4B-Instruct-2507-w4a16-2-2048/README.md | Documentation | 2.0 KB | — |
| Qwen3-4B-Instruct-2507-w4a16-2-2048/README_CN.md | Documentation | 1.9 KB | — |
| Qwen3-4B-Instruct-2507-w4a16-2-4096/README.md | Documentation | 2.0 KB | — |
| Qwen3-4B-Instruct-2507-w4a16-2-4096/README_CN.md | Documentation | 1.9 KB | — |
| Qwen3-4B-Instruct-2507-w8a8-2-1024/README.md | Documentation | 2.0 KB | — |
| Qwen3-4B-Instruct-2507-w8a8-2-1024/README_CN.md | Documentation | 1.9 KB | — |
| Qwen3-4B-Instruct-2507-w8a8-2-2048/README.md | Documentation | 2.0 KB | — |
| Qwen3-4B-Instruct-2507-w8a8-2-2048/README_CN.md | Documentation | 1.9 KB | — |
| Qwen3-4B-Instruct-2507-w8a8-2-4096/README.md | Documentation | 2.0 KB | — |
| Qwen3-4B-Instruct-2507-w8a8-2-4096/README_CN.md | Documentation | 1.9 KB | — |
| README.md | Documentation | 3.8 KB | — |
| README_CN.md | Documentation | 3.5 KB | — |
| Qwen3-4B-Instruct-2507-w4a16-2-1024/Qwen3-4B-Instruct-2507.rkllm | Other | 3.0 GB | aa59bf49920c |
| Qwen3-4B-Instruct-2507-w4a16-2-1024/SHA256SUMS | Other | 328 B | — |
| Qwen3-4B-Instruct-2507-w4a16-2-16384/Qwen3-4B-Instruct-2507.rkllm | Other | 3.4 GB | 6672f7944e7e |
| Qwen3-4B-Instruct-2507-w4a16-2-16384/SHA256SUMS | Other | 328 B | — |
| Qwen3-4B-Instruct-2507-w4a16-2-2048/Qwen3-4B-Instruct-2507.rkllm | Other | 3.0 GB | 0f2efdb6a1ab |
| Qwen3-4B-Instruct-2507-w4a16-2-2048/SHA256SUMS | Other | 328 B | — |
| Qwen3-4B-Instruct-2507-w4a16-2-4096/Qwen3-4B-Instruct-2507.rkllm | Other | 3.0 GB | 437efef80c6d |
| Qwen3-4B-Instruct-2507-w4a16-2-4096/SHA256SUMS | Other | 328 B | — |
| Qwen3-4B-Instruct-2507-w8a8-2-1024/Qwen3-4B-Instruct-2507.rkllm | Other | 4.8 GB | f4409cca98a5 |
| Qwen3-4B-Instruct-2507-w8a8-2-1024/SHA256SUMS | Other | 328 B | — |
| Qwen3-4B-Instruct-2507-w8a8-2-2048/Qwen3-4B-Instruct-2507.rkllm | Other | 4.8 GB | e84c898890ac |
| Qwen3-4B-Instruct-2507-w8a8-2-2048/SHA256SUMS | Other | 328 B | — |
| Qwen3-4B-Instruct-2507-w8a8-2-4096/Qwen3-4B-Instruct-2507.rkllm | Other | 4.9 GB | c0ff238ac8ed |
| Qwen3-4B-Instruct-2507-w8a8-2-4096/SHA256SUMS | Other | 328 B | — |
| .gitattributes | Repository | 2.2 KB | — |
License and Download
- License
- apache-2.0
- Access
- Open weights, no gate
Released by RKNNAI through its official repository on Hugging Face. Read the license.
Questions About RK3576-LLM-Qwen3-4B-Instruct-2507
Can I use RK3576-LLM-Qwen3-4B-Instruct-2507 commercially?
Yes. RK3576-LLM-Qwen3-4B-Instruct-2507 is released under Apache License 2.0. The Apache License 2.0 is a permissive open-source license. It permits commercial use, modification and redistribution. It requires keeping the license and copyright notices and any NOTICE file, stating significant changes, and it includes an express patent grant from contributors.