I have been anti-low quant. I can feel the difference, I swear. I am also anti-quantized KV cache. I have been burned there - I think the tradeoffs compound, especially at longer contexts. Due to this, I have been running Qwen 3.8 27B (UD-Q8_K_L) on 2x of the 3090s in TP. I swear by it. It is such a
