Series

A two-part series on running local language models on CPU-only homelab hardware: what the benchmarks actually say, and what context length does to your token rate.