How to Run ComfyUI for Local Image Generation
Install ComfyUI on Windows, Linux, or Mac, load a checkpoint, and generate your first image on port 8188 — exact commands and error strings.
Guides
Step-by-step tutorials: installing and running local LLMs, choosing hardware, building RAG pipelines and working with agents — written to be followed, not just read.
How to Run a Local LLM on Linux with Ollama
Install Ollama on Linux with one curl line, wire the systemd service, and get the GPU detected — exact "no compatible GPUs" log strings and fixes.
How to Run a Local LLM with llama.cpp
Install llama.cpp, serve a GGUF model with the unified llama CLI on port 9931, and benchmark with llama bench — exact commands and error strings.
How to Run a Local LLM on macOS with Ollama
Install Ollama on macOS (Sonoma+, Apple Silicon): drag Ollama.app, fix the Gatekeeper "damaged" error, pull llama3.1:8b, and check ~21 tok/s on M4.
How to Run a Local LLM Server with vLLM
Install vLLM with uv, serve any Hugging Face model with one command on port 8000, and read the exact CUDA driver errors before you hit them.
Best GPU for Local AI in 2026: Ranked by VRAM
Seven GPUs ranked for local AI, from a $200 P40 to a $5,000 RTX 5090. Prices as of October 2026, what each card runs, and which one to buy for your model.
Run Llama 3.1 8B on a 16 GB GPU
Install Ollama and run Llama 3.1 8B at 35-68 tokens per second on a 16 GB GPU: VRAM math, context limits, and the exact errors to fix.
Best 7B–8B LLMs for 8 GB VRAM: Qwen 3 8B, Gemma 3, Mistral
Qwen 3 8B, Gemma 3 12B, Mistral 7B and Llama 3.1 8B on 8 GB cards: exact Q4_K_M sizes, VRAM totals and tok/s on an RTX 4060. One fits, one is a trap.
How to Estimate VRAM for Any LLM: The Math
Two formulas — weight size and KV cache — predict the VRAM any LLM needs. Worked examples for 8B models on 8 GB and 16 GB GPUs.
LM Studio vs Ollama: Which Should You Use in 2026?
LM Studio 0.4 vs Ollama 0.40 compared: ports, GPU support, licenses, MCP, cloud. Both free — here is which one to install first and why.
What Is MCP? A Practical Guide to the Model Context Protocol
Set up MCP in Claude Desktop with npx: the exact config file, a working filesystem server, and the security rules that keep your machine safe.
How to Run a Local LLM on Windows with Ollama
Install Ollama on Windows in five steps: pull a 4.9 GB model, check ollama ps, and call the API on port 11434 - exact error strings included.
How to Choose a GPU for Local AI: VRAM First
Pick the model first, then the GPU: VRAM math for 8B to 70B models, card tiers with street prices as of October 2026, and what each tier runs.
Build a Local RAG Pipeline with Ollama and LlamaIndex
A fully local RAG stack in ~50 lines: nomic-embed-text embeddings, Chroma vector store, LlamaIndex retrieval, llama3.1:8b answers — no cloud, no API keys.