Qwen3-Embedding-0.6B — RKLLM v1.2.3 (w8a8, RK3588)
RKLLM conversion of Qwen/Qwen3-Embedding-0.6B for Rockchip RK3588 NPU inference.
Converted with RKLLM Toolkit v1.2.3. This model generates dense vector embeddings for semantic search, RAG pipelines, and similarity tasks — all running on the NPU without a GPU.
Key Details
| Property | Value |
|---|---|
| Base Model | Qwen/Qwen3-Embedding-0.6B |
| Toolkit Version | RKLLM Toolkit v1.2.3 |
| Runtime Version | RKLLM Runtime ≥ v1.2.1 (v1.2.3 recommended) |
| Quantization | w8a8 (8-bit weights, 8-bit activations) |
| Target Platform | RK3588 |
| NPU Cores | 3 |
| Optimization Level | 1 |
| Hybrid Ratio | 0.5 |
| Model Type | Embedding |
| Languages | English, Chinese (multilingual) |
Why This Model?
Running a dedicated embedding model on the RK3588 NPU allows the main LLM to use the full NPU without context-switching. Qwen3-Embedding-0.6B achieves strong retrieval performance at a compact size, making it ideal for local RAG pipelines on edge hardware.
Pair with the Qwen3-Reranker-0.6B for a complete retrieval stack.
Hardware Tested
- Orange Pi 5 Plus — RK3588, 16GB RAM, Armbian Linux
- RKNPU driver 0.9.8
- RKLLM Runtime v1.2.3
Usage
With the RKLLM Embedding Service
mkdir -p ~/models/Qwen3-Embedding-0.6B
cd ~/models/Qwen3-Embedding-0.6B
git lfs install && git clone https://huggingface.co/GatekeeperZA/Qwen3-Embedding-0.6B-RKLLM-v1.2.3 .
Use with GatekeeperZA/RKLLM-API-Server embedding endpoint.
File Listing
| File | Description |
|---|---|
Qwen3-Embedding-0.6B-rk3588-w8a8-opt-1-hybrid-ratio-0.5.rkllm |
Quantized embedding model for RK3588 NPU |
Compatibility Notes
- Minimum runtime: RKLLM Runtime v1.2.1. v1.2.3 recommended.
- RKNPU driver: ≥ 0.9.6
- SoCs: RK3588 / RK3588S. Not compatible with RK3576 without reconversion.
- RAM: ~1GB loaded.
Acknowledgements
- Alibaba Qwen Team for Qwen3-Embedding
- Rockchip / airockchip for the RKLLM toolkit and runtime
- Converted by GatekeeperZA