Is Together AI Down?

No — Together AI is up

Reachable from all 8 checked regions

Average response time: 209ms

Last checked 3 hours ago · checks run every 6 hours

Official status page: https://status.together.ai

Together AI uptime

100%

Last 7 days

100%

Last 30 days

100%

Last 90 days

406ms

Avg response, 30 days

Measured from multiple regions every 6 hours. Percentages count only checks that returned an availability answer — 44 days measured so far. A dash means that window does not yet hold enough measured days to publish a figure.

30-day history

44-day clean streak

Sep 10 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.

ams

302ms

DNS 37ms TCP 2ms TLS 9ms TTFB 302ms

arn

307ms

DNS 76ms TCP 1ms TLS 8ms TTFB 307ms

cdg

223ms

DNS 34ms TCP 1ms TLS 8ms TTFB 223ms

dfw

136ms

DNS 42ms TCP 2ms TLS 8ms TTFB 135ms

sin

249ms

DNS 11ms TCP 3ms TLS 12ms TTFB 248ms

sjc

89ms

DNS 35ms TCP 3ms TLS 7ms TTFB 89ms

syd

213ms

DNS 38ms TCP 1ms TLS 7ms TTFB 213ms

yyz

155ms

DNS 57ms TCP 1ms TLS 8ms TTFB 155ms

What Together AI does

Together AI runs open source and open weight models as a hosted API. Developers send chat, vision, image, embedding and speech requests to endpoints that are OpenAI-compatible and pay per token instead of running GPUs themselves. It also offers fine-tuning and dedicated endpoints, which makes it a direct dependency for applications that route their model calls through it.

What an outage looks like

Calls to api.together.xyz return 5xx errors or time out, and applications waiting on a model response stall rather than failing cleanly. More often the failure is narrower than that: one model returns errors or answers far more slowly than usual while every other model on the platform is fine. The Playground and marketing site may be unaffected throughout.

What to do about it

Check status.together.ai, which reports individual models as their own components under Inference groups for Chat, Vision, Images, Embeddings and Voice. Find your specific model rather than reading the overall banner. If your model is the affected one, switching to another of similar capability is usually faster than waiting, which is the practical benefit of an OpenAI-compatible API.

Is it down for everyone, or just you?

If this page says Together AI 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 17 cities across three regions to find out where it breaks.

Test it yourself

Related services

Together AI outage FAQ

Is Together AI down, or is it just my model?

Usually just the model. Together AI's status page breaks out individual models as separate components, with Chat, Vision, Images, Embeddings and Voice each listing the specific models underneath. One model degrading while the platform stays healthy is the common case, and it is the distinction crowd-report aggregators cannot make.

Will my in-flight requests complete?

Requests already accepted may still return, but one that times out has no queue behind it on a synchronous inference call. It is simply gone, and retrying is your job. Use backoff rather than a tight retry loop, since retrying hard against a degraded endpoint adds to the load that caused the problem in the first place.

Does an incident affect fine-tuning jobs as well as inference?

The status page will not tell you directly. It tracks Website, Playground and the inference model groups, and does not break fine-tuning out as its own component. Check a running job through the API rather than assuming an inference incident stopped it, and rather than assuming it did not.

Should I fail over to another provider?

If your application tolerates a different model's output, yes. The API is OpenAI-compatible, so routing to another provider or another Together model is mostly a base URL and model name change. Applications with prompts tuned to one specific model get less benefit from this, which is a good argument for testing a fallback model before you actually need it.

How we measure this

We request Together AI'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 Together AI 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 Together AI goes down

This page refreshes every 6 hours. Your own monitors run as often as every 30 seconds, from up to 17 cities across three regions, and tell you the moment something breaks.

Start Free Monitoring

Free plan available No credit card required