qwen3.6 performance jump is real, just make sure you have it properly configured

I've been running workloads that I typically only trust Opus and Codex with, and I can confirm 3.6 is really capable. Of