AI workloads / Public reference

AI API latency

A supplied apiLatencyMs or routingLatencyMs of 500 or more.

What you need to supply

Telemetry you include, in milliseconds.

If it reports a finding

Medium below 2,000 ms, and high at or above that. The suggested next step is to inspect routing, timeouts, model choice, and region.

What this does not prove

It uses the number you supply. It does not measure the provider itself.

Telemetry field

{ "apiLatencyMs": 500 }

All review checks. Reading this page does not submit your infrastructure.