I've been using DS4 flash as a local "big brain" for when my qwen 3.6 35b a3b gets stuck on something, and for planning. I'm blown away by how capable it is at a 2 bit quant Has anyone used the GGUF at a similar 2bit quant and compared performance? I've got a Nimo Strix Halo box, 128GB, kyuz0 stack