Is Cohere Down?

No — Cohere is up

Reachable from all 8 checked regions

Average response time: 840ms

Last checked 2 hours ago · checks run every 6 hours

Official status page: https://status.cohere.com

Cohere uptime

100%

Last 7 days

100%

Last 30 days

100%

Last 90 days

1077ms

Avg response, 30 days

Measured from multiple regions every 6 hours. Percentages count only checks that returned an availability answer — 25 days measured so far. A dash means that window does not yet hold enough measured days to publish a figure.

30-day history

25-day clean streak

Aug 22 Today

No downtime Partial Downtime Not measurable No data

Reachability by region

Each region runs its own request from a different part of the world. A service can be up for one continent and down for another, which is usually the first sign of a routing or CDN problem.

bom

943ms

DNS 4ms TCP 1ms TLS 9ms TTFB 941ms

cdg

620ms

DNS 72ms TCP 9ms TLS 18ms TTFB 611ms

dfw

906ms

DNS 39ms TCP 0ms TLS 53ms TTFB 904ms

ewr

522ms

DNS 17ms TCP 23ms TLS 19ms TTFB 520ms

fra

699ms

DNS 41ms TCP 1ms TLS 9ms TTFB 608ms

gru

1350ms

DNS 227ms TCP 1ms TLS 10ms TTFB 1348ms

iad

487ms

DNS 50ms TCP 2ms TLS 8ms TTFB 479ms

jnb

1198ms

DNS 46ms TCP 0ms TLS 59ms TTFB 1196ms

What Cohere does

Cohere builds large language models sold as an API for enterprises, covering text generation, embeddings, reranking and retrieval-augmented search. It targets companies that want models deployed in their own cloud or a private environment rather than only as a public API, and it sits underneath search, classification and assistant features built on top of it.

What an outage looks like

API calls return 5xx responses or take far longer than usual to complete, and features built on Cohere such as semantic search, reranking and chat assistants either fail or fall back to whatever your code does when a model call errors. The dashboard at dashboard.cohere.com may fail to load. Degraded latency across endpoints is a more common pattern here than a hard outage.

What to do about it

Check status.cohere.com, which separates Endpoints, Models, Infrastructure, Playground, Docs and Website and lists 25 models individually, so confirm whether the model you call is the affected one. Watch the name collision: status.cohere.io belongs to a different company. If reranking is the degraded piece, serving unranked results is usually better than serving none.

Is it down for everyone, or just you?

If this page says Cohere is up but it is not loading for you, the problem is between you and them. Run a check against any URL from up to 18 cities across three regions to find out where it breaks.

Test it yourself

Related services

Cohere outage FAQ

Which Cohere is this page about?

Cohere the AI company at cohere.com, which sells large language models to enterprises. Two other companies share the name: Cohere at cohere.io, a customer support cobrowsing product with its own status page at status.cohere.io, and Cohere Health at coherehealth.com. A search for cohere status returns all three, which is worth knowing before you read the wrong page.

Is the whole platform down or just one model?

Often just one model. Cohere's status page groups 25 models as individual components beneath a Models heading, kept separate from Endpoints and Infrastructure. A specific model degrading while the API itself is healthy is a different incident from the API being unreachable, and only the per-component view tells them apart.

What do Cohere incidents usually look like?

Slower rather than broken. Its status page carries a published incident titled Degraded performance across endpoints and models, which is the shape most Cohere incidents take. Degraded latency is harder to spot from inside an application than an outright error, because calls still succeed but take long enough to blow through user-facing timeouts.

What should my application do when Cohere is slow?

Set an explicit timeout on model calls and decide the fallback before you need it. Degraded latency is the common failure mode here, and without a timeout a slow Cohere call becomes a slow page for your own users. A cached result, a smaller model or a plain message all beat a request that hangs.

How we measure this

We request Cohere's public endpoint every 6 hours from Fly.io regions across six continents — the most recent check ran from 8 of them.
A region counts as down only when it gets no usable HTTP response. A 403 or 429 means the origin answered and refused us, which we report as blocked, never as an outage.
A single failing region is treated as probe noise. We only change the verdict when two consecutive cycles agree.
Response times average only the regions that actually served the page, so a timeout never inflates the number.
Where Cohere publishes an official status feed we read it too. An all-clear from the vendor can soften an unconfirmed degradation; a vendor-declared outage only worsens our verdict when our own checks corroborate it.

Get alerted when Cohere goes down

This page refreshes every 6 hours. Your own monitors run as often as every 30 seconds, from up to 18 cities across three regions, and tell you the moment something breaks.

Start Free Monitoring

Free plan available No credit card required