Skip to content

Anthropic outage history and postmortems

AI · San Francisco, US · 93 incidents in 90 days · 97.4% implied uptime · last 3d ago

Status page Compare

Incidents

400 on file
StartedIncidentDuration
Sep 2200:57 UTCElevated errors for multiple models1h 38m
Sep 1616:18 UTCIssues with Google Play subscriptions33m
Sep 1509:48 UTCIntermittent error spikes for Claude Mythos 5.1 and Claude Fable 5.11h 26m
Sep 1114:19 UTCElevated errors for Claude Mythos 5.1 and Claude Fable 5.117m
Sep 1021:43 UTCElevated latency on the Claude API for some users1h 41m
Sep 1015:54 UTCDegraded functionality for Claude Cowork on Windows4d 3h
Sep 313:26 UTCElevated errors for multiple models2h 57m
Sep 312:37 UTCElevated errors for Claude Sonnet 519m
Sep 221:17 UTCElevated errors for Claude Sonnet 526m
Sep 123:26 UTCDelays in credit purchases1h 58m
Sep 117:05 UTCDegraded performance on platform.claude.com and Claude for Microsoft Office 3651h 2m
Sep 116:02 UTCDegraded performance on Claude for Microsoft Office 36520m
Aug 3119:21 UTCDegraded performance on claude.ai and Claude Code1h 14m
Aug 3117:23 UTCDegraded performance on claude.ai29m
Aug 2817:22 UTCElevated errors on Claude Code and Claude Cowork2h 59m
Aug 2420:11 UTCIssues logging into Claude.ai15m
Aug 2417:12 UTCErrors logging into Claude.ai3m
Aug 2405:06 UTCElevated errors for multiple models3h 24m
Aug 2019:16 UTCElevated errors on requests to multiple models26m
Aug 2018:32 UTCElevated errors on Google connectors29m
Aug 1909:42 UTCDegraded performance for Claude Opus 5 and Claude Haiku 4.51h 20m
Aug 1816:20 UTCDegraded performance for multiple models2h 41m
Aug 1713:56 UTCDegraded performance for Claude Opus 5, Claude Sonnet 51h 33m
Aug 1621:58 UTCService disruption on Claude services36m
Aug 1423:57 UTCElevated errors for Claude Fable 530m
Aug 1420:30 UTCService disruption on Claude services1h 28m
Aug 1407:58 UTCIssues reaching status.claude.com2h 55m
Aug 1400:36 UTCService disruption on Claude Code on the web and related services15m
Aug 1314:33 UTCElevated errors for Claude Mythos 5, Claude Fable 5, and Claude Sonnet 51h 35m
Aug 1213:50 UTCDegraded performance for multiple models4h 17m
Aug 513:51 UTCDegraded performance for Claude Opus 543m
Aug 507:05 UTCDegraded performance of multiple models7h 9m
Aug 420:48 UTCElevated errors across many models1h 11m
Aug 411:27 UTCElevated errors on Claude Sonnet 525m
Aug 406:16 UTCElevated errors on Claude Sonnet 539m
Aug 315:13 UTCDegraded performance on Claude Sonnet 515m
Aug 312:52 UTCError rates across multiple models1h 24m
Jul 3106:18 UTCDegraded performance on Claude Sonnet 547m
Jul 3013:43 UTCDegraded performance on Claude Opus 4.841m
Jul 3005:57 UTCElevated errors across many models4h 51m
About Anthropic

Anthropic, PBC is an American artificial intelligence (AI) public benefit corporation headquartered in San Francisco, California. Its flagship product is Claude, a series of proprietary large language models (LLMs).

The last year, day by day

No incident Minor Major CriticalDeeper colour: longer outage

Downtime, added up

AI median
2d 9h

Anthropic: 2d 9h weighted downtime in range, category median 5h 8m.

93 incidents, 31 major or critical, 9d 17h of incident time (6d 2h major or critical, 3d 14h partial), longest 4d 3h, implied uptime 97.4%, between Jun 28, 2026 and Sep 25, 2026.

Across fru.dev:Company profileAcquisitionsComplianceFree tiersLeadersboardPaydaysChangefeedSLAsTechConfUptime

Questions

Is Anthropic down right now?

Check Anthropic's own status page at https://status.claude.com for the live state. This page shows Anthropic's incident history from that page; the most recent incident started 2026-09-22: "Elevated errors for multiple models".

How often does Anthropic have outages?

Anthropic posted 94 incidents in the last 90 days and 400 in the last year, 31 of them major or critical in the last 90 days. The median incident lasted 1h.

When was Anthropic's last major outage?

2026-09-22: "Elevated errors for multiple models", lasting 1h 38m.

What usually causes Anthropic incidents?

Anthropic's status updates in this period do not name a root cause.

Sources: vendors' own status pages, published postmortems and SEC 8-K Item 1.05 filings, read daily. Times as reported. Logos via logo.dev; trademarks belong to their owners.

Outages by email

Saturday mornings: the week's major outages, new postmortems and disclosed breaches, only in weeks that had some.

Double opt-in. Unsubscribe any time.