Decoder. plain-English AI glossary

P99 latency

● Core

The latency at the 99th percentile — 99% of requests are faster, 1% are slower.

Think of it like

If you survey 100 people, p99 is the slowest person's time.

Example

P99 latency = 200ms means 99 out of 100 requests complete in 200ms or less; 1 in 100 takes longer.

How it actually works

P99 is a standard SLA target in production. It's one number that sums up the tail. Other percentiles: p95 (95th), p50 (median), p999 (99.9th). P99 is strict enough to catch problems but not so strict (like p999) that outliers dominate.

For product teams

P99 latency is your SLA promise to customers — publish it and defend it.

For engineers

Set P99 budgets and profile to find bottlenecks; reduce variance more than average.

Related

Read anything AI without the jargon

Look up any term in plain English, or save terms as you read with the free Chrome extension.

Open DecoderAdd to Chrome