Qwen 3.8 what am I missing here?
retained text
I'm just starting my journey here into local LLM. Since I work in VFX I've got a 5090 and 256GB of sys mem. Installed LM studio and downloaded Qwen 3.8 27B since that very comfortably fits in VRAM. Set it to low effort and all the other typical starting defaults. I ask it to give me a three line poem about space. Comes back in a couple seconds. I ask it to change from space to a forest. No sweat. I ask it to give me Houdini VEX code to delete half the points in a particle simulation. Pretty fast. I ask it to download and install the various VFX skills out there on the net. Still cool. I ask if it can work inside of Houdini for me and it explains the MCP thing and that it can check for a connection when Houdini is up and running. So I boot up the Houdini app and say "Houdini is up and running now" Qwen now takes minutes just to process that very simple prompt. Then it starts reasoning/thinking with my 5090 at 100% for minutes upon minutes and no UI updates at all. Eventually I just kill the process. Where am I going wrong? All you folks seem to be using Qwen for huge tasks and all kinds of shit and I can't even seem to get it to reply to me saying I just launched an app.
document fetched 2026-10-31 00:45 · canonical https://reddit.com/r/LocalLLaMA/comments/1xj2hi1/qwen_38_what_am_i_missing_here