MoE Models for Local LLMs: Why a 30B Model Can Run Like a 3B One
MoE models for local LLMs explained: total vs active parameters, why speed follows active weights, memory needs of gpt-oss and Qwen3-30B-A3B, and offload tips.
MoE models for local LLMs explained: total vs active parameters, why speed follows active weights, memory needs of gpt-oss and Qwen3-30B-A3B, and offload tips.
MoE 模型本地运行指南:总参数与激活参数的区别,速度为何取决于激活权重,gpt-oss 和 Qwen3-30B-A3B 的内存需求,以及 CPU 卸载技巧。
gpt-oss local deployment with Ollama and llama.cpp: official memory needs for 20b and 120b, MXFP4, MoE offload flags, and reasoning effort.
gpt-oss 本地部署教程:用 Ollama 与 llama.cpp 运行 20b 和 120b,对照官方内存需求、MXFP4 量化、MoE 卸载参数与推理强度设置。
Best local LLM for your hardware: picks by 8GB, 12GB, 24GB and 32GB VRAM and by task, covering Qwen, Gemma and gpt-oss, with file sizes and VRAM estimates.
本地大模型推荐:按 8GB、12GB、24GB、32GB 显存和聊天、编程、推理、看图等用途,给出 Qwen、Gemma、gpt-oss 等选择,附文件大小与显存估算。