ParentFull threadmrkn1·No reason other than their Q4 model working reasonably well and fast on my CPU laptop. Should work with any ONNX VLM modelView on HN