$1800 (in GPU cost running with P2P running Qwen/Qwen3.6-27b-FP8 with 262K context and BF16 KV cache at 55 tok/s(reddit.com)