Two weeks fine-tuning 96GB VRAM for local LLMs, paid APIs triumphed.

I run a homelab with four RTX 3090s — 96 GB of VRAM, 44 CPU cores. For two weeks I tried to make it my daily driver for local LLM inference instead of paying for cloud APIs. I got it working. Then I looked at the numbers and subscribed to a paid API anyway. Here's the uncomfortable part, and the opt