User shares best settings for Qwen 3.6 27B on 48GB VRAM

A Reddit user details their optimal configuration for running Qwen 3.6 27B (Q8_0) on dual RTX 4090 + RTX 3090 with 48GB VRAM using llama.cpp and tensor split. The post invites community feedback on the setup.