Local Qwen with 5090 chokes for minutes on a trivial prompt once Houdini MCP connects
Paddy_Tanninger · reddit · 2026-10-30
A VFX professional running a local 27B-class Qwen model via LM Studio on an RTX 5090 with 256GB RAM found simple tasks (poems, rewrites, Houdini VEX code generation) fast and smooth, and the model correctly explained how to connect via MCP.
But once Houdini was actually launched and he told the model "Houdini is up and running now," it took minutes just to process that trivial prompt — reasoning at 100% GPU load with no UI updates until he killed the process. He's asking the community what he's doing wrong, highlighting a real performance pitfall when local models handle MCP tool-calling contexts.
More from Models
- DeepSeek V4.1 Flash tops open-model usage among research agents, index data shows — AkariAsai · 2026-10-30
- Dev claims Gemini Argon makes 'researcher-level jumps' other models miss as 'carbon' looms — vitaliychiley · 2026-10-30
- Meta's Alexandr Wang touts Muse: 6.7x more concise than ChatGPT at answering the same question — alexandr_wang · 2026-10-30
- OpenAI's subsidy shift rewrites the score-vs-cost Pareto frontier, analysis finds — RexDouglass · 2026-10-30
- Phonic tops VAmoS benchmark for agentic speech-to-speech models — graceisford · 2026-10-30
- DeepSeek Frontend Leaks cyber-v3/v4 Keys; Still Hiring Attackers as V4 Excels at Security — teortaxesTex · 2026-10-30