Qwen 3.8 27B on MBP
Hey everyone, I typically read and learn from everyone else in here, but wanted to get some thoughts on reasonable expectations running Qwen 3.8-27B on a M5 Max, 128 GB RAM. I've been running frontier models via Codex, Claude Code, Hermes and some other harnesses for quite some time have done quite a bit of development with them. However, I'm new to local models. My hunch is that my impatience or lack of awe with the local models has more to do with me being use to running frontier models vs. watching the open-weight models evolve over time. So, what are some reasonable expectations, given the hardware I have, for the complexity of tasks I can hand to a model like Qwen 3.8, given a comparable task the speed difference between a frontier vs. open-weight and compared to running on Nvidia hardware vs. Apple, is there a huge difference in throughput. I hope what I'm asking isn't too vague or ambiguous as I'm intrigued by the possibilities of running local, but I'm not quite feeling like I can hand it anything heavier than scripts and configurations at the moment. Thanks in advance for any thoughts.