Metric
Claude
OpenAI
Mistral
Grok
DeepSeek
Kimi
Cohere
- Reliability
- Composite reliability score50.758.870.088.387.658.3▲89.8
- Uptime98.677%99.643%99.915%100.000%99.922%98.869%100.000%
- Total downtime9h 31m2h 34m37m0m34m8h 8m0m
- Active right now0000000
- Composite reliability score
- Incidents
- Total incidents463613151401
- Major + critical184101400
- Minor27281305000
- Longest single incident5h 44m32h 4m29h 33m10h 2m▲34m1h 35m2h 54m
- Longest clean streak3d 15hMay 22–266dMay 13–193d 5hJun 3–78d 3hMay 18–2629d 14hMay 8 – Jun 78d 1hMay 14–22▲30dMay 8 – Jun 7
- Total incidents
- Time to fix
- Median fix time39m1h 31m5m1h 44m34m▲3m2h 54m
- On a bad day3h 9m13h 2m▲19m10h 2m34m37m2h 54m
- Closed vs still open46 closedall wrapped up35 closedall wrapped up131 closedall wrapped up5 closedall wrapped up1 closedall wrapped up40 closedall wrapped up1 closedall wrapped up
- Median fix time
- Patterns
- What broke mostClaude Code39 incidents · 18h 47m attributedConversations13 incidents · 32h 7m attributedConversations API35 incidents · 5h 53m attributedAPI Console2 incidents · 11h 46m attributedAPI 服务 (API Service)1 incident · 34m attributedAgentic Model23 incidents · 8h 11m attributedUnspecified1 incident · 2h 54m attributed
- Most common failure typeElevated errors38 of 46Elevated errors13 of 36Degraded95 of 131Degraded2 of 5Other1 of 1Other40 of 40Other1 of 1
- Busiest day of weekFriday11 incidents on this dayFriday9 incidents on this dayMonday36 incidents on this dayWednesday2 incidents on this dayToo few to callFriday15 incidents on this dayToo few to call
- Busiest hour of day8 AM UTC5 incidents started here5 PM UTC5 incidents started here4 PM UTC17 incidents started here3 PM UTC2 incidents started hereToo few to call9 AM UTC8 incidents started hereToo few to call
- What broke most
- No synthetic API probing — we only see what each provider publishes on their Statuspage. Quietly degraded responses that never get acknowledged on the status page are invisible here.
- No regional breakdown — Statuspage feeds don't reliably expose region-level impact, so a US-only outage and a global one count the same.
- No request-volume weighting — “1 incident” means one acknowledged incident, regardless of how many calls it actually affected.