Service status

Uptime isn't an adjective, it's a record you can check

These figures come from real probes (currently every 15 minutes; the interval is stated in the uptime commitment), not a hand-maintained table. A hand-maintained status page will always forget to update on the day something breaks — the day everyone is looking at it.

Some capability tiers have no real upstream yet, so their probes hit a local stub — those figures reflect the stub's health, not service availability. Tiers that are connected are probed against the real upstream.

All systems operational

OperationalDegradedDown

Open incidents — No incidents currently open. Post-mortems

frontier-a

Frontier
No data yetOutside CN only
Operational
90 days agoToday

frontier-b

Frontier
No data yetOutside CN only
Operational
90 days agoToday

balanced-a

Balanced
No data yetOutside CN only
Operational
90 days agoToday

balanced-b

Balanced
No data yetOutside CN only
Operational
90 days agoToday

balanced-c

Balanced
No data yetOutside CN only
Operational
90 days agoToday

domestic-a

Domestic flagship
No data yetAvailable in CN
Operational
90 days agoToday

domestic-b

Domestic flagship
No data yetAvailable in CN
Operational
90 days agoToday

domestic-c

Domestic flagship
No data yetAvailable in CN
Operational
90 days agoToday

fast-a

Fast & light
No data yetOutside CN only
Operational
90 days agoToday

fast-b

Fast & light
No data yetOutside CN only
Operational
90 days agoToday

fast-c

Fast & light
No data yetOutside CN only
Operational
90 days agoToday

fast-d

Fast & light
No data yetOutside CN only
Operational
90 days agoToday

Uptime is computed from **outage windows**, not from a probe success rate: two consecutive failed probes open an outage, dated from the first failure, closing at the next success. Probes run every 15 minutes **from our own node — there are no multi-region probe points yet**, so this figure reflects availability as seen from where we sit. History is retained for 90 days. Planned maintenance is announced in advance and excluded from the figure — but we still mark it on this page.

Raw probe records

The commitment says raw records are published, not just the rolled-up percentage. This is that record: every probe's timestamp, outcome, latency and failure kind. Uptime is computed from these rows — you can recompute it yourself.

frontier24/24Export a full month of raw records (CSV):2026-102026-092026-082026-07
TimeResultLatencyFailure
10/2 09:12● ok2722ms
10/2 08:57● ok2751ms
10/2 08:42● ok2691ms
10/2 08:27● ok2784ms
10/2 08:12● ok3080ms
10/2 07:57● ok2372ms
10/2 07:42● ok3071ms
10/2 07:27● ok2477ms
10/2 07:12● ok2846ms
10/2 06:57● ok3195ms
10/2 06:42● ok2492ms
10/2 06:27● ok2580ms
10/2 06:12● ok2692ms
10/2 05:57● ok2520ms
10/2 05:42● ok2623ms
10/2 05:27● ok4666ms
10/2 05:12● ok2710ms
10/2 04:57● ok2693ms
10/2 04:42● ok4400ms
10/2 04:27● ok3032ms
10/2 04:12● ok2702ms
10/2 03:57● ok5082ms
10/2 03:42● ok2688ms
10/2 03:27● ok3011ms
balanced24/24Export a full month of raw records (CSV):2026-102026-092026-082026-07
TimeResultLatencyFailure
10/2 09:12● ok2347ms
10/2 08:57● ok2421ms
10/2 08:42● ok2345ms
10/2 08:27● ok2180ms
10/2 08:12● ok3592ms
10/2 07:57● ok4037ms
10/2 07:42● ok2579ms
10/2 07:27● ok2601ms
10/2 07:12● ok2579ms
10/2 06:57● ok4251ms
10/2 06:42● ok2616ms
10/2 06:27● ok2978ms
10/2 06:12● ok2771ms
10/2 05:57● ok3315ms
10/2 05:42● ok2690ms
10/2 05:27● ok2419ms
10/2 05:12● ok2949ms
10/2 04:57● ok2238ms
10/2 04:42● ok4403ms
10/2 04:27● ok3169ms
10/2 04:12● ok4238ms
10/2 03:57● ok2750ms
10/2 03:42● ok3443ms
10/2 03:27● ok3411ms
domestic24/24Export a full month of raw records (CSV):2026-102026-092026-082026-07
TimeResultLatencyFailure
10/2 09:12● ok639ms
10/2 08:57● ok525ms
10/2 08:42● ok455ms
10/2 08:27● ok1795ms
10/2 08:12● ok760ms
10/2 07:57● ok437ms
10/2 07:42● ok1677ms
10/2 07:27● ok422ms
10/2 07:12● ok537ms
10/2 06:57● ok566ms
10/2 06:42● ok446ms
10/2 06:27● ok552ms
10/2 06:12● ok3620ms
10/2 05:57● ok837ms
10/2 05:42● ok628ms
10/2 05:27● ok3679ms
10/2 05:12● ok4716ms
10/2 04:57● ok531ms
10/2 04:42● ok452ms
10/2 04:27● ok531ms
10/2 04:12● ok1844ms
10/2 03:57● ok30849ms
10/2 03:42● ok25724ms
10/2 03:27● ok643ms
fast24/24Export a full month of raw records (CSV):2026-102026-092026-082026-07
TimeResultLatencyFailure
10/2 09:12● ok0ms
10/2 08:57● ok0ms
10/2 08:42● ok0ms
10/2 08:27● ok0ms
10/2 08:12● ok0ms
10/2 07:57● ok0ms
10/2 07:42● ok0ms
10/2 07:27● ok0ms
10/2 07:12● ok0ms
10/2 06:57● ok0ms
10/2 06:42● ok0ms
10/2 06:27● ok0ms
10/2 06:12● ok0ms
10/2 05:57● ok0ms
10/2 05:42● ok0ms
10/2 05:27● ok0ms
10/2 05:12● ok0ms
10/2 04:57● ok0ms
10/2 04:42● ok0ms
10/2 04:27● ok0ms
10/2 04:12● ok0ms
10/2 03:57● ok0ms
10/2 03:42● ok0ms
10/2 03:27● ok0ms

Raw records are kept 90 days. In a dispute, export the month and check it yourself — no need to contact us.

Get notified

You shouldn't have to refresh this page to find out something is wrong.

Email

One when it opens, one when it clears

Sent only when a tier you actually use is affected. No marketing, and unsubscribing takes effect immediately.

Webhook

Push it to your own alerting

Structured JSON with affected models, tiers, estimated impact and an incident ID — drop it straight into your on-call flow.