Latency

Latency is the delay between sending a request and getting a response started. It's a number that shapes how good an AI product feels to use — voice conversations in particular only feel natural when latency approaches human conversational reaction time (0.2-0.3 seconds).

"Smarter" and "faster" pull in opposite directions — reasoning models get smarter by thinking longer, but that makes them slower. That's why companies release fast, small models alongside deep, large ones, and new chip announcements usually boast about how much they've cut latency.