RunAIHome is a practical resource for people who want to run AI on their own hardware. Not enterprise GPUs. Not cloud APIs you can't audit. Just a consumer card, an OS you control, and software you understand.

Why this site exists

The local AI scene moves fast and most coverage either lives in scattered Reddit threads or in breathless news posts that never get to the actual numbers. We do the slow, boring work: repeatable benchmarks, end-to-end tutorials, and side-by-side comparisons that survive past the next driver update.

What you'll find here

• GPU benchmarks for AI workloads — Stable Diffusion, ComfyUI, Flux, local LLM inference, training on limited VRAM.
• Step-by-step tutorials — Ollama, llama.cpp, ComfyUI workflows, LoRA training, quantization.
• Honest tool comparisons — Ollama vs LM Studio, ComfyUI vs A1111, cloud vs local for specific workloads.
• Self-hosted AI stack — Open WebUI, n8n automations, privacy-first setups.

Editorial approach

Every benchmark you read here was run on real hardware we own, with the exact configuration documented. Every tutorial was followed end-to-end before publishing. Affiliate links exist where they make sense (you can usually tell), but they never decide which tool we recommend.

How We Test

Every benchmark and recommendation on this site comes from hands-on testing. Our primary test rigs:

• Primary GPU rig: RTX 4090 24GB, AMD Ryzen 9 7950X, 128GB DDR5, Windows 11 Pro
• Mid-range rig: RTX 3090 24GB, Intel i9-13900K, 64GB DDR4
• Apple Silicon: Mac Studio M3 Ultra 192GB, MacBook Pro M4 Pro 48GB
• AMD test: RX 7900 XTX 24GB, ROCm testing on Ubuntu 24.04 and Windows 11

Models are tested with Ollama, llama.cpp, vLLM, and LM Studio. Image generation tested with ComfyUI and Automatic1111. All numbers are our own measurements unless explicitly cited.

Contact

Questions, corrections, or hardware you'd like us to test? Reach out via the contact email listed on our GitHub or wait for the contact page coming soon.