Is the Cerebras API down? Live status, latency & uptime by region
· rebuilt about every 30 minutes
Is the Cerebras API down right now?
Reachable from all 4 regions, requests to gpt-oss-120b succeeding
Last successful probe Oct 22, 03:46 UTC · every 5 minutes from 4 region(s) · status feed checked Oct 22, 03:17 UTC
3 isolated failure(s) in the last 72 hours, the most recent (requests to gpt-oss-120b) visible from only one region (Europe (Germany)). A single-region failure cannot be separated from a problem on that network path, so it is not called an outage. Measured from outside cerebras’s own network, independently of its status page; full log: measured AI API incidents.
Real requests to gpt-oss-120b: 0 of 1 succeeded in the last 24 hours from 1 region(s); last failure Oct 21, 10:26 UTC from Europe (Germany) ([Errno 101] Network is unreachable). A one-token completion is sent on the same schedule as the host probe; 5xx, 529 and timeouts count as failures, problems with our own key or quota are excluded.
Cerebras’s status page lists no open incident (checked Oct 22, 03:17 UTC).
As of October 22, 2026, the cerebras AI API responds in 76 ms p50 (edge, time-to-first-byte) from US (Central) — its fastest measured region — with 100% uptime. Slowest is Asia (Tokyo) at 205 ms p50.
Endpoint response (TTFB), 24 h
76 ms
Typical (p50) from US (Central), the fastest region
Uptime, 24 h
99.8%
1,152 probes across 4 region(s)
Failed probes, 72 h
2
of 3,459 · Europe (Germany), South America (São Paulo)
Failed requests, 24 h
1
of 1 to gpt-oss-120b
Reported by Cerebras, 72 h
0
on status.cerebras.ai
Was Cerebras down in the last 3 days?
No. 3 isolated single-region probe failure(s) were recorded, none of which meets the bar for an outage (two or more regions at once).
requests to gpt-oss-120b: Europe (Germany) only — not an outage · under one 5-minute probe cycle · 1 failed probe(s) · [Errno 101] Network is unreachable
Europe (Germany) only — not an outage · at least 15 minutes · 2 failed probe(s) · [Errno 101] Network is unreachable
requests to gpt-oss-120b: South America (São Paulo) only — not an outage · under one 5-minute probe cycle · 1 failed probe(s) · timeout
Earlier days with Cerebras incidents: Sep 25 (2 outages) · Aug 27 (status page) · Jul 29 (status page) · Jul 28 (1 outage)
Nothing. Cerebras’s status page lists no incident in the last 72 hours (checked Oct 22, 03:17 UTC).
Is Cerebras getting faster or slower?
Endpoint response (TTFB) · milliseconds · lower is faster
Last 72 hours · UTC · newest time on the right · current hour is incomplete.
Each line shows a region’s typical response per hour (p50), using the same scale. A gap means no successful timed probe in that hour.
typical response per hour (p50)all probes succeededsome failed (under half)half or more failedno probes recorded
Probe counts by region
Europe (Germany): 2 failed / 864 endpoint probesEurope (Germany) · model: 1 failed / 1 requests · 71 hour(s) without dataUS (Central): 0 failed / 862 endpoint probesAsia (Tokyo): 0 failed / 862 endpoint probesSouth America (São Paulo): 0 failed / 862 endpoint probesSouth America (São Paulo) · model: 1 failed / 1 requests · 71 hour(s) without data
Chart details
The host strip shows endpoint probe outcomes; the model strip shows completion-request outcomes for gpt-oss-120b. The model strip does not show time to first token (TTFT). Adjacent hours with all successes or no data are joined. Hover an hour for its recorded values.
Latest is the most recent hourly p50 with data. Median is the median of hourly p50 values over these 72 hours. The “slower than usual” label means latest is at least 25% above that median; it is a relative change, not a fixed speed rating.
How fast is the Cerebras API from each region?
Endpoint response (TTFB) · last 24 hours · milliseconds (ms) · lower is faster.
How to read these metrics
Endpoint response (TTFB)Network connection and the first response byte, without model inference.
Typical · p5050% of successful timed samples finish within this time.
Slower · p9595% of successful timed samples finish within this time.
Region
Typical (p50 TTFB)
Slower (p95)
Uptime
Asia (Tokyo)
205 ms
283 ms
100%
Europe (Germany)
198 ms
302 ms
99.3%
South America (São Paulo)
192 ms
258 ms
100%
US (Central)
76 ms
151 ms
100%
Measured from US (Central), cerebras responds in 76 ms; from Asia (Tokyo) the same API takes 205 ms — 2.7× longer for an identical request. That difference is the network path, not the model, and it applies to every call you make.
What is actually being measured here?
A probe in each region opens a real connection to cerebras’s own API host every five
minutes and times DNS resolution, the TCP handshake, the TLS handshake and the first byte of the
response. Nothing is routed through a gateway or aggregator, so these figures describe
cerebras’s infrastructure rather than a reseller’s. Uptime counts a probe as failed only when the
service genuinely fails — a 401 from an unauthenticated probe means the endpoint answered
correctly and is counted as up.
What this page does not tell you
These are edge latency numbers: the time before the model begins generating. They
say nothing about answer quality, throughput, price, or how long a full completion takes — those
depend on the model you call. Edge latency matters because it is unavoidable: it is paid on every
request, before a single token exists, and no prompt engineering removes it. Where an API key is
available, inference time-to-first-token is measured separately and shown on the region pages with the
model named next to the figure. Likewise, an unauthenticated probe cannot see errors that hit only some
models or some accounts — the kind a status page usually reports — which is why Cerebras’s
own reports are shown next to the measurements rather than folded into them.
— embed code