CheckUpstream measured Replicate at 100.000% availability across 2821 independent checks (global · api). 16 incidents were recorded in the past 12 months; the most recent resolved 21 days ago after 15 minutes. Status signals are polled from Replicate's official incident.io status page, alongside independent probes and community reports.
90 days measured availability how this is computed
100.000%
global · api · n=2821 · p95 532ms (30d)
Share of our own checks against Replicate endpoints that returned Replicate's expected response. Evidence that their API edge is answering correctly — not that every product behind it works.
Vendor disclosed: 2 incidents · 90 of 90 days with none posted · 11m of declared impact
Fewer incidents in the last 30 days than 18% of 232 tracked services.
Share of each window with no impact declared by Replicate
7d
100.000%
30d
99.972%
90d
99.991%
180d
99.995%
1y
99.998%
Downtime like this can be refundable
If you're a paying Replicate customer, outages in these windows may count against Replicate's SLA — vendors owe service credits for missed uptime targets, but only if you file inside the deadline. CheckUpstream tracks the entitlement, drafts the claim, and hands you the filing link.
Track your Replicate entitlement →Does Replicate actually matter to you?
A Replicate outage only matters if something you ship depends on it. Point us at a public repository and we will read its dependencies and tell you — no account, nothing stored.
Check a repository →Aug 10, 21:05 UTC · resolved in 14m · Resolved
Aug 2, 16:09 UTC · resolved in 10m · Resolved
Live Replicate status for your README or docs. Updates itself — no API key, no maintenance.
[](https://checkupstream.com/reliability/replicate)GitHub READMEs, GitLab, MDX docs — links to the live reliability page.
Replicate is currently Operational. CheckUpstream tracks Replicate in real-time across the official status page, RSS / Atom feeds, and community telemetry. Refresh this page for the latest reading.
Looking for the live "is Replicate down" view? See the Replicate Live status page.
Replicate's last resolved incident started 21 days ago and lasted 14 minutes (degraded performance).
Replicate's 30-day measured availability is 100.000%, from CheckUpstream's own probes against Replicate endpoints. It is the share of checks that returned the expected response, not a figure taken from Replicate's status page.
Sign up to CheckUpstream and add Replicate to a watchlist. We send incident alerts via Slack, email, webhook, or PagerDuty within seconds of detection, covering both the official status page and community signals (HN, Reddit, Bluesky, Cloudflare Radar).
We poll Replicate's official status page on a 1-5 minute cadence, parse its RSS / Atom feed as a fallback, ingest community signals from social platforms, and aggregate per-org SDK telemetry. Multiple sources reduce false positives versus any single feed alone.