Real benchmarks. Real hardware. Real costs.
Run AI at home, practically.
GPU benchmarks, local LLM tutorials, ComfyUI workflows, and self-hosted AI guides for the home lab crowd. Tested on the cards you can actually buy.
What we cover
Four corners of the home AI lab.
-
GPU Benchmarks
Real numbers for Stable Diffusion, Flux, ComfyUI, and local LLMs across the cards normal humans can afford.
-
Local LLM Tutorials
Step-by-step Ollama, llama.cpp, LM Studio, and quantized model picks that fit your VRAM, not a datacenter's.
-
ComfyUI & Image Gen
Workflows that work, LoRA training on limited VRAM, and node setups without the YouTube guesswork.
-
Self-Hosted Stack
Open WebUI, n8n automations, privacy-first infra. Your data, your hardware, your rules.
Latest posts
Fresh from the bench.
-
HIP Out of Memory on AMD? Every Fix for PyTorch, ComfyUI, llama.cpp, and Strix Halo (2026)
Fix 'HIP out of memory' on AMD GPUs and Ryzen AI Max APUs: the PYTORCH_HIP_ALLOC_CONF flag, ComfyUI --reserve-vram, and the GTT limit that hides your RAM.
Read guide → -
Strata Runs a 125B Model on a 12GB Gaming GPU: Real Benchmarks, the 64GB RAM Catch, and Who Should Install It (2026)
Strata runs Qwen3.8-Flash-Next 125B on a 12GB gaming GPU at 53-94 tok/s. Verified benchmarks, RAM and SSD requirements, and who should actually use it.
Read guide → -
Aleph Alpha Kolibri-1 for Local AI in 2026: The 78B Apache 2.0 MoE Your Tools Can't Load Yet
Kolibri-1 local AI hardware guide: real GGUF sizes, measured tok/s on CPU and a 48GB Mac, the llama.cpp patch you need, and which hardware (if any) to buy.
Read guide → -
FlashAttention Only Supports Ampere GPUs or Newer? Every Fix for Turing, Volta, and Pascal Cards (2026)
Fix the 'FlashAttention only supports Ampere GPUs or newer' error in vLLM and Transformers — dtype flags, SDPA fallback, and the llama.cpp path for older GPUs.
Read guide → -
NVIDIA's AI Agent Watchdog Is Not a Chip for Your RTX Card: What Sentry, OpenShell, and BlueField-4 Mean for Home Lab Agent Rigs in 2026
NVIDIA's Sentry watchdog runs on BlueField-4 DPUs, not consumer GPUs. What home lab builders running local AI agents should actually do about it in 2026.
Read guide →