ai
Is Replicate down right now?
Our independent multi-region reading is: reachable.
Independently measured by StackWitness. We ping each service host ourselves (reachability) and also read its status page. The strong signal is when a host is unreachable while the vendor claims operational.
Current status
reachable = the service host answered our probe; not a guarantee the product is fully healthy
Incident timeline
No independently measured incidents in this window.
Probe latency (this week)
Probe round-trip time and reachability outcomes only. Not full product health or a vendor SLA.
7 disruptions · baseline ~101ms p50 · scale fits this week’s range (not a fixed 0-100ms axis)
Line / fill = successful probe p50 · Bands = disruption windows · Floor ticks = fail share in bucket
- Elevated probe latency (p50 189ms, p95 485ms vs baseline ~101ms) between 2026-09-05T03:01:09.364Z and 2026-09-05T09:01:09.364Z. Probe RTT only, not full product health.
- Elevated probe latency spanning 2026-09-05T15:01:09.364Z → 2026-09-06T09:01:09.364Z (peak p95 ~555ms). Probe RTT only.
- Elevated probe latency spanning 2026-09-06T21:01:09.364Z → 2026-09-07T09:01:09.364Z (peak p95 ~546ms). Probe RTT only.
- Elevated probe latency spanning 2026-09-07T21:01:09.364Z → 2026-09-08T09:01:09.364Z (peak p95 ~546ms). Probe RTT only.
- Elevated probe latency spanning 2026-09-08T21:01:09.364Z → 2026-09-09T09:01:09.364Z (peak p95 ~548ms). Probe RTT only.
- Elevated probe latency (p50 189ms, p95 466ms vs baseline ~101ms) between 2026-09-10T03:01:09.364Z and 2026-09-10T09:01:09.364Z. Probe RTT only, not full product health.
- Elevated probe latency (p50 182ms, p95 509ms vs baseline ~101ms) between 2026-09-11T03:01:09.364Z and 2026-09-11T09:01:09.364Z. Probe RTT only, not full product health.
How to read this page
StackWitness shows two columns for Replicate: independent multi-region reachability (our probes of the service host) and the vendor's own status page (self-report). Divergence matters most when we cannot reach a host the vendor still calls operational. Reachable means the host answered our probe; it is not a promise every feature works, and we never issue a compliance verdict.
Questions
Is Replicate down?
StackWitness independently pings the Replicate service host for reachability from multiple regions and also reads the vendor status page. Reachable means the host answered our probe; it is not a claim that every product feature is healthy. We never issue a compliance verdict.
How does StackWitness know if Replicate is down?
We ping the Replicate service host for reachability from multiple regions and read the vendor status page. The signal worth acting on is when we cannot reach a host the vendor still calls operational.
What does "reachable" mean here?
The host answered our probe from our regions. It is not a claim every Replicate feature is healthy, and we never issue a compliance verdict.
Measurement window: → .
We independently ping this service host (reachability) and also read its vendor status page. The strong signal is when we cannot reach a host that still claims to be operational. We never invent calm green or issue a compliance verdict.