Some monitored APIs are degraded
Updated Sat, 31 Oct 2026 00:00:00 UTC · regions: ap-south-1, us-east-1, eu-west-1
Providers
Anthropicclaude-haiku-5-5412 msOperational
100% of the last 24 hours fully operational · p95 780 ms
- ap-south-1 · 350 ms TTFT
- us-east-1 · 424 ms TTFT
- eu-west-1 · 499 ms TTFT
OpenAIsample-small-model388 msOperational
95.8% of the last 24 hours fully operational · p95 720 ms
- ap-south-1 · 330 ms TTFT
- us-east-1 · 400 ms TTFT
- eu-west-1 · 469 ms TTFT
Google Geminisample-small-model301 msOperational
95.8% of the last 24 hours fully operational · p95 610 ms
- ap-south-1 · 256 ms TTFT
- us-east-1 · 310 ms TTFT
- eu-west-1 · 364 ms TTFT
Mistralsample-small-model455 msDegraded
79.2% of the last 24 hours fully operational · p95 1180 ms
- ap-south-1 · 387 ms TTFT
- us-east-1 · 469 ms TTFT
- eu-west-1 · 551 ms TTFT
Recent incidents
- Mistral: Elevated latency from US East and EU WestInvestigating
Started Fri, 30 Oct 2026 19:00:00 UTC
- Google Gemini: Brief spike in time to first tokenResolved
Started Fri, 30 Oct 2026 12:00:00 UTC · resolved Fri, 30 Oct 2026 13:00:00 UTC
Each point on the history bars is one hour. A failed probe counts against a provider only when the failure is on the provider's side; rate limits and credential problems on our own account are tracked separately and never shown as outages.