Show HN: I tested 15 free AI models at building real software on a $25/year VPS

Build a standardized, automated benchmarking harness that allows developers to stress-test various open-source models on low-cost, resource-constrained hardware. This tool would quantify the real-world utility of different model weights for specific software engineering tasks under strict memory and compute constraints.

Suggested repo: vps-bench

"Stop guessing which LLM runs on your $25 VPS: The definitive benchmark for budget AI development."

Estimated effort: 40h