Latency is the delay between a caller's speech or action and the voice system's response.

Why Latency matters

Excessive delay makes conversation feel unnatural, encourages callers to repeat themselves, and can create accidental interruptions or abandoned calls.

How it works in practice

End-to-end latency includes audio transport, speech processing, reasoning, tool calls, and speech generation, so teams should measure the full experience rather than one model in isolation.