Reference · Glossary

Load testing

Simulating **realistic or peak traffic** against an API to measure latency, errors, and resource limits before launch.

#When to use

Pre-launch validation, autoscaling tuning, and proving rate limits hold under stress.

Uncontrolled tests against production without caps — you can trigger provider bans and real cost.

Ramp `/api/chat` from 10 to 500 rps while watching p95 latency and token spend.