What's the best planning LLM for 100GB+ RAM/VRAM local setups? Reddit weighs in
GodComplecs · reddit · 2026-10-17
A Reddit user asks which local model is best for planning tasks (science/research first, coding for drafting plans) on 100GB+ RAM/VRAM hardware at 10+ tokens/s, soliciting what others actually run.
More from Infra
- AI agents are having their Kubernetes moment: agent logic sinks into the platform layer — bibryam · 2026-10-17
- Why inference providers must choose: fast (Groq) or cheap (CoreWeave), not both — AccBalanced · 2026-10-17
- Dev who tried Mac Studio for local LLMs: 'I can run it at home != I can use it for work' — TheZachMueller · 2026-10-17
- Credo's ALC optical cable targets a short-reach optics pool ~5x its $13B copper market by 2030 — BenBajarin · 2026-10-17
- SemiAnalysis: Neoclouds adding AMD GPUs or TPUs refused entry to GTC Berlin — AccBalanced · 2026-10-17
- vLLM gets autonomously optimized by AI agents with clear per-step performance gains, a first — vllm_project · 2026-10-17