Vital signs for the AI APIs you depend on.
Probegnosys probes leading LLM APIs from several regions and publishes what it actually measures, not what a provider's own status page chooses to say.
Current status
Sample data, last 24 hoursReachable is not the same as working.
A provider can return a page and still be failing your requests. We measure the request path end to end, from three regions, and tell you which part broke.
Where the time goes
DNS, connect, TLS, first token and total time, recorded separately so a slow response has an address.
Their outage, not ours
Rate limits, auth errors and timeouts are told apart, so our own problems never get reported as theirs.
Same call, three vantage points
A provider can be fine in one region and struggling in another. We show both.
How it works
Probe
A small, real request goes to each provider from every region on a fixed five-minute schedule.
Classify
Each result is sorted into a fixed set of outcomes, separating provider faults from probe-side faults.
Publish
The status page is rebuilt from the latest results, with history you can check, not just a green dot.
GPU cloud price and availability
The same probe platform, pointed at GPU capacity: what is on offer, at what price, and whether you can actually get it.