Live reliability of 7 major AI providers, pulled straight from their official status feeds. No synthetic probes — what you see is what they reported.
Every public incident from each provider's status feed, 90 days back. Which component broke, how often, how long to fix.
One number, 0–100, weighted Uptime 50% · Frequency 20% · Severity 20% · MTTR 10%. Critical outages count more than partials; many small blips never beat one sustained critical.
Status pages only show yesterday and tend to flatter the provider. We show 90 days side-by-side, with the math shown, so you stop guessing whose outage broke your app.