Qwen-3.8-9B-nf4-cw-ov

This is an OpenVINO NF4 weight-only quantized (Compressed Weights) version of empero-ai/Qwen3.8-9B-Distill, optimized for Intel NPU acceleration and deployment via OpenVINO Model Server (OVMS).

Model Details

• Base Model: empero-ai/Qwen3.8-9B-Distill
• Quantization: NF4 Compressed Weights (via Intel NNCF)
• Target Hardware: Intel NPU 4 (Lunar Lake Architecture and newer)
• Deployment Framework: OpenVINO Model Server (OVMS) / OpenVINO C++ Runtime

Compatibility & System Requirements

Hardware Requirements

• NPU Generation: Intel NPU 4 (found in Intel Core Ultra 200V series / Lunar Lake) or newer.
• Note: Legacy NPUs (NPU 3 / Meteor Lake) or older iGPU/CPU-only environments may not support this specific NF4 weight-compression layout without fallback to CPU runtime.

For detailed architecture info, evaluation benchmarks, and original weights, refer to the base repository: **[empero-ai/Qwen3.8-9B-Distill](https://huggingface.co/empero-ai/Qwen3.8-9B-Distill

More techinical detail

Reccomend Enviroment and techinical detail:My blog(※Japanese)