qwen3-4b-int4-ov

OpenVINO IR export of Qwen/Qwen3-4B, quantized to INT4 — 2180 MB.

Quantization	weight compression, group 128, symmetric, AWQ + scale estimation
Lesson	08 LLM chatbot / 09 speculative main / 12 RAG
Built by	convert/convert_all.py of the ARCademy OpenVINO courseware

openvino_genai.LLMPipeline(model_dir, device). Append ' /no_think' to a user turn to skip the reasoning block.

from huggingface_hub import snapshot_download
model_dir = snapshot_download("circulus/qwen3-4b-int4-ov")