Model providers
One model can be served by several upstream providers. See how they differ and how to find which one served a call.
On this page
What a provider is
A provider is the upstream that actually runs inference. One model can be served by several providers — the vendor's own API and cloud-hosted versions, for example. Same model, different regions, limits and reliability.
How providers differ
Latencyfirst tokenRegion and queueing mean first-token latency for the same model can vary noticeably between providers.
Reliabilityrecent recordThe gateway keeps watching each provider's success rate and moves recently weak ones down or skips them for a while.
CapacityconcurrencyEach provider can only take so many requests at once; when one is full the gateway moves to another.
CapabilitiesdifferencesSome capabilities exist only on some providers; requests only go to providers that support what the call needs.
Finding out which one served a call
Search the request ID (the X-Maxx-Request-ID response header) in console › Usage › Request log to see the call's status, usage, first-token latency and the model that finally served it.
You never need to choose a provider — routing does it, and switches automatically when one fails. See failover.
Contact us
Something unclear, or need a human? Pick the channel that fits your question.
Supportsupport@apione.ai
WeCom
Telegram
Communityhttps://t.me/apione
Support tickets
Billing & accountConsole › Feedback