I’m building AI voice agents for my startup on Vapi and got annoyed never knowing which LLM was actually fastest. so I built a tiny free tool, that tracks TTFT across models + gives you a simple API to always pick the fastest one. https://metrik-dashboard.vercel.app/