Metrik – Real-time LLM latency for voice agents and free API

I built Metrik to answer a simple question for AI voice agents: which LLM is fastest right now?

If you use Vapi with multiple providers (OpenAI, Anthropic, Google, etc.), it’s hard to: – Measure Time to First Token (TTFT) consistently – See when a provider slows down – Route calls to the fastest model without manual changes

What Metrik does – Continuously pings multiple LLMs and logs TTFT/latency – Shows a live dashboard by provider/model – Can route Vapi agents to the currently fastest allowed model – Exposes a free API so you can pull the metrics into your own tools

I’d love feedback on what other metrics you care about (besides TTFT), and whether you’d want this as a hosted service, self-hosted, or just a library.