HomeIncidents
Data API

AI API incident history — OpenAI, Anthropic, Google and 40+ APIs, measured from 4 regions

In the last 72 hours 11 outage(s) could be attributed to a provider with confidence. Most recent: yi-01ai failed from 2 of 4 regions at once on 2026-10-23 14:44 UTC, for at least 4 minutes, while no other provider was failing.

Independent third-party outage log for AI inference APIs, built from probes run directly against each provider’s endpoint rather than from status-page polling. Observation started 2026-07-25 (90 days so far); the raw series is retained for 90 days, and the incident window below is kept deliberately shorter than that — see the last section for why.

Multi-region failures

Last 72 hours · UTC · newest time on the right.

One row per provider. Each block spans a recorded failure in two or more regions; block width shows its duration. Single-region failures remain in the tables below.

provider outage: 2+ regions, no other provider failingshared failure: cause not attributed

Multi-region failures by provider, last 72 hoursEach row is a provider. Block width shows failure duration. Provider outages are separate from multi-region failures shared with other providers, whose cause is not attributed. Exact UTC times, durations and classifications are listed in the tables below.Oct 21Oct 22Oct 23nowgoogle: Oct 23, 16:59 – 17:35 UTC · 4 of 4 regions · at least 37 minutes · 2 other provider(s) failed in the same minutes — not attributedgoogle: Oct 23, 15:35 – 15:47 UTC · 3 of 4 regions · at least 12 minutes · 1 other provider(s) failed in the same minutes — not attributedgoogle: Oct 23, 13:55 – 14:27 UTC · 4 of 4 regions · at least 32 minutes · 2 other provider(s) failed in the same minutes — not attributedgoogle: Oct 23, 13:19 – 13:27 UTC · 2 of 4 regions · at least 9 minutes · provider outage — nobody else was failinggoogle: Oct 22, 19:55 – 20:21 UTC · 3 of 4 regions · at least 26 minutes · 1 other provider(s) failed in the same minutes — not attributedgoogle: Oct 22, 18:55 – 19:06 UTC · 3 of 4 regions · at least 10 minutes · provider outage — nobody else was failinggoogle: Oct 22, 15:21 – 17:41 UTC · 4 of 4 regions · at least 140 minutes · 5 other provider(s) failed in the same minutes — not attributedgoogle: Oct 22, 14:41 – 14:46 UTC · 3 of 4 regions · at least 5 minutes · provider outage — nobody else was failinggoogle: Oct 22, 12:35 – 14:06 UTC · 4 of 4 regions · at least 90 minutes · 2 other provider(s) failed in the same minutes — not attributedgoogle: Oct 22, 12:01 – 12:01 UTC · 2 of 4 regions · under one 5-minute probe cycle · provider outage — nobody else was failinggoogle: Oct 22, 06:41 – 07:01 UTC · 3 of 4 regions · at least 20 minutes · 1 other provider(s) failed in the same minutes — not attributedgoogle: Oct 21, 16:01 – 16:01 UTC · 2 of 4 regions · under one 5-minute probe cycle · provider outage — nobody else was failinggoogle: Oct 21, 15:41 – 15:41 UTC · 2 of 4 regions · under one 5-minute probe cycle · provider outage — nobody else was failinggoogle: Oct 21, 15:21 – 15:21 UTC · 2 of 4 regions · under one 5-minute probe cycle · 1 other provider(s) failed in the same minutes — not attributedgoogle: Oct 21, 15:00 – 15:01 UTC · 4 of 4 regions · under one 5-minute probe cycle · 1 other provider(s) failed in the same minutes — not attributedgoogle: Oct 21, 14:40 – 14:41 UTC · 4 of 4 regions · under one 5-minute probe cycle · provider outage — nobody else was failinggoogle: Oct 21, 14:20 – 14:21 UTC · 4 of 4 regions · under one 5-minute probe cycle · 2 other provider(s) failed in the same minutes — not attributedgoogle: Oct 21, 14:00 – 14:01 UTC · 4 of 4 regions · under one 5-minute probe cycle · provider outage — nobody else was failinggoogle: Oct 21, 08:00 – 08:01 UTC · 4 of 4 regions · under one 5-minute probe cycle · 1 other provider(s) failed in the same minutes — not attributedyi-01ai: Oct 23, 14:44 – 14:47 UTC · 2 of 4 regions · at least 4 minutes · provider outage — nobody else was failingyi-01ai: Oct 23, 07:47 – 07:50 UTC · 2 of 4 regions · at least 3 minutes · provider outage — nobody else was failingdoubao: Oct 23, 09:30 – 09:45 UTC · 4 of 4 regions · at least 14 minutes · 1 other provider(s) failed in the same minutes — not attributedwriter: Oct 22, 15:40 – 15:41 UTC · 3 of 4 regions · under one 5-minute probe cycle · 2 other provider(s) failed in the same minutes — not attributedfriendli: Oct 21, 00:55 – 01:01 UTC · 3 of 4 regions · at least 6 minutes · provider outage — nobody else was failing
googleyi-01aidoubaowriterfriendli

How do you know it was the provider and not my network?

Every provider is probed from 4 independent regions (Asia (Tokyo), Europe (Germany), South America (São Paulo), US (Central)) every five minutes, and all 46 of them are probed on the same schedule. Two questions follow from that, and an incident has to survive both.

Did it fail from more than one place? If a request fails from one region while the others succeed, the likely cause is the path between that region and the provider — a peering problem, a route change, one bad edge node. We do not call that an outage.

Was it the only thing failing? If several unrelated APIs stop answering in the same few minutes, the shared cause sits above all of them: a transit route, a CDN both sit behind, or our own probes. Blaming one provider for that would be guessing. Those events are listed separately below and attributed to nobody. What is left — failing from several regions at once, alone — is the provider, and when its own host answers 5xx rather than going silent, the provider has said so itself. Neither a status page nor a crowd-reporting site can make this distinction: one sees only what the provider admits, the other only what users happen to report.

Which AI APIs had confirmed outages in the last 72 hours?

Started (UTC)ProviderRegions hitWhichDurationFailed probesDid they say so?
2026-10-23 14:44yi-01ai2 of 4Asia (Tokyo), Europe (Germany)≥4 min2No machine-readable status page
2026-10-23 13:19google
requests to gemini-flash-lite-latest		2 of 4Europe (Germany), South America (São Paulo)≥9 min2No machine-readable status page
2026-10-23 07:47yi-01ai2 of 4Asia (Tokyo), US (Central)≥3 min2No machine-readable status page
2026-10-22 18:55google
requests to gemini-flash-lite-latest		3 of 4Asia (Tokyo), South America (São Paulo), US (Central)≥10 min3No machine-readable status page
2026-10-22 14:41google
requests to gemini-flash-lite-latest		3 of 4Asia (Tokyo), Europe (Germany), South America (São Paulo)≥5 min3No machine-readable status page
2026-10-22 12:01google
requests to gemini-flash-lite-latest		2 of 4Asia (Tokyo), Europe (Germany)<5 min2No machine-readable status page
2026-10-21 16:01google
requests to gemini-flash-lite-latest		2 of 4Europe (Germany), South America (São Paulo)<5 min2No machine-readable status page
2026-10-21 15:41google
requests to gemini-flash-lite-latest		2 of 4Europe (Germany), South America (São Paulo)<5 min2No machine-readable status page
2026-10-21 14:40google
requests to gemini-flash-lite-latest		4 of 4Asia (Tokyo), Europe (Germany), South America (São Paulo), US (Central)<5 min4No machine-readable status page
2026-10-21 14:00google
requests to gemini-flash-lite-latest		4 of 4Asia (Tokyo), Europe (Germany), South America (São Paulo), US (Central)<5 min4No machine-readable status page
2026-10-21 00:55friendli3 of 4Europe (Germany), South America (São Paulo), US (Central)≥6 min6No machine-readable status page

Was an AI API down on a specific day?

One page per UTC day since observation started, with what the probes measured and what each provider reported about itself.

44 earlier day(s) are linked from each day page.

Did the provider admit it?

The last column checks each failure against the provider’s own status page. Only 11 of the 46 providers tracked here publish a machine-readable status feed at all; for the rest there is nothing to check against, which is itself worth knowing before you plan around one.

Where a feed exists, both outcomes are informative. When it carries a matching entry, two independent records agree and the timing shows which noticed first. When it carries none, we recorded a failure their page does not mention — and that is not an accusation. Status pages have thresholds, and a short burst of 5xx can sit legitimately below one. We report the discrepancy; explaining it is the provider’s to do.

Which failures could not be attributed to anyone?

These hit several regions, but other providers were failing in the same minutes, so a shared upstream cause is at least as likely as a coincidence of separate outages. They are published because omitting them would flatter the data, and left unattributed because that is what the evidence supports.

Started (UTC)ProviderRegions hitWhichDurationFailed probesDid they say so?
2026-10-23 16:59google
requests to gemini-flash-lite-latest		4 of 4Asia (Tokyo), Europe (Germany), South America (São Paulo), US (Central)≥37 min7No machine-readable status page
2026-10-23 15:35google
requests to gemini-flash-lite-latest		3 of 4Europe (Germany), South America (São Paulo), US (Central)≥12 min3No machine-readable status page
2026-10-23 13:55google
requests to gemini-flash-lite-latest		4 of 4Asia (Tokyo), Europe (Germany), South America (São Paulo), US (Central)≥32 min7No machine-readable status page
2026-10-23 09:30doubao4 of 4Asia (Tokyo), Europe (Germany), South America (São Paulo), US (Central)≥14 min9No machine-readable status page
2026-10-22 19:55google
requests to gemini-flash-lite-latest		3 of 4Asia (Tokyo), South America (São Paulo), US (Central)≥26 min4No machine-readable status page
2026-10-22 15:40writer3 of 4Europe (Germany), South America (São Paulo), US (Central)<5 min3No machine-readable status page
2026-10-22 15:21google
requests to gemini-flash-lite-latest		4 of 4Asia (Tokyo), Europe (Germany), South America (São Paulo), US (Central)≥140 min26No machine-readable status page
2026-10-22 12:35google
requests to gemini-flash-lite-latest		4 of 4Asia (Tokyo), Europe (Germany), South America (São Paulo), US (Central)≥90 min18No machine-readable status page
2026-10-22 06:41google
requests to gemini-flash-lite-latest		3 of 4Asia (Tokyo), Europe (Germany), US (Central)≥20 min5No machine-readable status page
2026-10-21 15:21google
requests to gemini-flash-lite-latest		2 of 4Asia (Tokyo), South America (São Paulo)<5 min2No machine-readable status page
2026-10-21 15:00google
requests to gemini-flash-lite-latest		4 of 4Asia (Tokyo), Europe (Germany), South America (São Paulo), US (Central)<5 min4No machine-readable status page
2026-10-21 14:20google
requests to gemini-flash-lite-latest		4 of 4Asia (Tokyo), Europe (Germany), South America (São Paulo), US (Central)<5 min4No machine-readable status page
2026-10-21 08:00google
requests to gemini-flash-lite-latest		4 of 4Asia (Tokyo), Europe (Germany), South America (São Paulo), US (Central)<5 min4No machine-readable status page

What isolated failures were recorded?

These failed from a single region only. They are published because hiding them would misrepresent the data, but they are not evidence that the provider was down.

Started (UTC)ProviderRegions hitWhichDurationFailed probesDid they say so?
2026-10-23 22:47yi-01ai1 of 4Asia (Tokyo)<5 min1No machine-readable status page
2026-10-23 22:32minimax1 of 4South America (São Paulo)<5 min1No machine-readable status page
2026-10-23 21:58google
requests to gemini-flash-lite-latest		1 of 4Europe (Germany)<5 min1No machine-readable status page
2026-10-23 19:24writer1 of 4Europe (Germany)<5 min1No machine-readable status page
2026-10-23 18:52yi-01ai1 of 4Asia (Tokyo)≥10 min2No machine-readable status page
2026-10-23 18:22yi-01ai1 of 4Asia (Tokyo)≥10 min2No machine-readable status page
2026-10-23 17:42sambanova1 of 4South America (São Paulo)<5 min1No entry in their status feed
2026-10-23 17:12yi-01ai1 of 4Asia (Tokyo)<5 min1No machine-readable status page
2026-10-23 16:55featherless1 of 4US (Central)<5 min1No machine-readable status page
2026-10-23 16:47yi-01ai1 of 4Asia (Tokyo)≥5 min2No machine-readable status page
2026-10-23 15:44iflytek1 of 4Europe (Germany)<5 min1No machine-readable status page
2026-10-23 15:25featherless1 of 4US (Central)<5 min1No machine-readable status page
2026-10-23 15:22yi-01ai1 of 4Asia (Tokyo)≥5 min2No machine-readable status page
2026-10-23 15:19google
requests to gemini-flash-lite-latest		1 of 4Europe (Germany)<5 min1No machine-readable status page
2026-10-23 15:02google
requests to gemini-flash-lite-latest		1 of 4Asia (Tokyo)<5 min1No machine-readable status page
2026-10-23 14:12yi-01ai1 of 4Asia (Tokyo)≥10 min3No machine-readable status page
2026-10-23 13:57yi-01ai1 of 4Asia (Tokyo)<5 min1No machine-readable status page
2026-10-23 13:54iflytek1 of 4Europe (Germany)<5 min1No machine-readable status page
2026-10-23 13:49baichuan1 of 4Europe (Germany)<5 min1No machine-readable status page
2026-10-23 13:47stepfun1 of 4Asia (Tokyo)<5 min1No machine-readable status page
2026-10-23 13:04writer1 of 4Europe (Germany)<5 min1No machine-readable status page
2026-10-23 12:57yi-01ai1 of 4Asia (Tokyo)≥5 min2No machine-readable status page
2026-10-23 12:02google
requests to gemini-flash-lite-latest		1 of 4Asia (Tokyo)<5 min1No machine-readable status page
2026-10-23 10:47yi-01ai1 of 4Asia (Tokyo)≥45 min7No machine-readable status page
2026-10-23 10:07yi-01ai1 of 4Asia (Tokyo)≥15 min3No machine-readable status page
2026-10-23 09:52yi-01ai1 of 4Asia (Tokyo)<5 min1No machine-readable status page
2026-10-23 09:27yi-01ai1 of 4Asia (Tokyo)≥5 min2No machine-readable status page
2026-10-23 08:57yi-01ai1 of 4Asia (Tokyo)≥10 min2No machine-readable status page
2026-10-23 08:12yi-01ai1 of 4Asia (Tokyo)≥5 min2No machine-readable status page
2026-10-23 06:52yi-01ai1 of 4Asia (Tokyo)<5 min1No machine-readable status page
2026-10-23 06:32doubao1 of 4Asia (Tokyo)<5 min1No machine-readable status page
2026-10-23 05:46yi-01ai1 of 4Asia (Tokyo)≥15 min3No machine-readable status page
2026-10-23 04:31yi-01ai1 of 4Asia (Tokyo)≥25 min4No machine-readable status page
2026-10-23 04:21google
requests to gemini-flash-lite-latest		1 of 4Asia (Tokyo)<5 min1No machine-readable status page
2026-10-23 03:56yi-01ai1 of 4Asia (Tokyo)<5 min1No machine-readable status page
2026-10-23 03:41featherless1 of 4Asia (Tokyo)<5 min1No machine-readable status page
2026-10-23 02:41google
requests to gemini-flash-lite-latest		1 of 4Asia (Tokyo)<5 min1No machine-readable status page
2026-10-23 02:11featherless1 of 4Asia (Tokyo)<5 min1No machine-readable status page
2026-10-23 02:06yi-01ai1 of 4Asia (Tokyo)≥10 min2No machine-readable status page
2026-10-23 00:21featherless1 of 4Asia (Tokyo)<5 min1No machine-readable status page

What did the providers’ own status pages report?

Every entry that the 11 providers with a machine-readable status feed opened in the same 72-hour window — self-reported, not verified by our probes, and shown so the two records can be read side by side. Most of these concern specific models or products: an unauthenticated probe of the API host cannot see them, and that is the gap between “the host answers” and “my requests succeed”. Feeds checked Oct 23, 23:17 UTC.

What counts as a failure here?

A probe fails when the connection times out, is refused, or the API host answers with a 5xx status. An authentication error does not count: the probes are unauthenticated by design, so a 401 or 403 means the endpoint answered correctly and is recorded as up. Latency alone — even very high latency — is not recorded as an incident; it appears in the p95 figures instead.

Why is this window only 72 hours?

This page focuses on the latest 72 hours. Raw measurements are retained for 90 days; earlier events are available in the daily incident archive above.

Outage and deprecation alerts use the same measurement record.

Method and limitations: methodology · Machine-readable: incidents.json · Licence: CC BY 4.0