ViewTube

What hardware do you actually need to run AI locally? I measure it and give you the numbers.

I test open models (Llama, Qwen, Gemma, DeepSeek) on real hardware — GPUs across price tiers, Apple Silicon, laptops and mini-PCs — with Ollama, LM Studio, llama.cpp and vLLM. Tokens per second, real VRAM use, what you lose when you quantise, power draw and noise. European prices. No marketing benchmarks.

The other half of the channel: building local agents that do real work — reading your documents, automating tasks, running without a paid API.

Local means your data never leaves your machine, you don't pay per token, and you don't depend on a company changing the price or the model overnight.

If you're about to buy hardware, or you want to get your AI out of the cloud, you're in the right place.