Is Replicate Down?
Live Replicate status. Auto-refreshes every 2 minutes.
Replicate is Operational
Replicate is up and running normally. All systems are operational.
Official status page: status.replicate.com
Component Status
Recent Replicate incidents
What to do when Replicate is down
Run the same models on another host
Most Replicate models are open weights wrapped in Cog, so they are not exclusive to Replicate. If the platform is down, you can run comparable image, audio, and language models through sibling inference gateways that host overlapping catalogs, or route across providers automatically so a single outage does not stop your pipeline. Hit the model owner's own API directly when one exists.
Switch media generation tools
If you were using Replicate specifically for image, video, or audio generation, comparable hosted media tools can cover the gap. Stability AI serves image models, Runway and Luma cover video, and ElevenLabs handles voice and audio, each on independent infrastructure from Replicate.
Get notified when status changes
Do not sit refreshing this page. Set an alert and get pinged the moment Replicate flips between operational, degraded, and down, so you can fail over before users notice. You can also watch every major AI provider on one board and compare open-weights model availability across hosts.
Frequently Asked Questions
Is Replicate down right now?
The live status indicator at the top of this page is the answer. A green operational dot means Replicate's API and model runs are responding normally. An amber dot means degraded performance, such as slow predictions, long queue times, or intermittent errors, and a red dot means an active outage. We refresh that signal against Replicate's own data, so read the dot first, then check the FAQs below if you need next steps.
How do I check if Replicate is down?
Start with the live indicator on this page, which reflects Replicate's official status. For the primary source, open status.replicate.com, where Replicate publishes component-level status and any active incidents. You can also confirm on your end by checking the Replicate dashboard for stuck or failed predictions, and watch for HTTP 5xx responses or unusually long cold-start times from the API.
What should I do when Replicate is down?
First confirm whether the issue is the API, a specific model, or a webhook delay, since a single model can fail while the platform is healthy. Retry with backoff for transient errors and check that your prediction is not just queued behind a cold start. If the outage is platform-wide, route image and video work to comparable hosted model providers or run the open weights you were using on another inference host until Replicate recovers.
How often does Replicate go down, and how reliable is it?
Replicate is generally reliable for a platform that runs thousands of community and first-party models, but it is not immune to incidents. The most common disruptions are not full outages: they are elevated cold-start latency, queue backups during demand spikes, and individual models failing or being deprecated. Treat the official status page as authoritative for actual uptime rather than relying on any single number.
Which models and services does Replicate run, and what gets confused for an outage?
Replicate hosts image models (such as the Flux and SDXL families), video models, audio and speech models, and language models, all packaged in the open-source Cog format and called through one API. A frequent false alarm is a cold start: if a model has not run recently, the first request spins up its container and can take many seconds or longer, which looks like a hang but is expected. Another is a model-specific error or a removed version, which affects only that model and not Replicate as a whole.
Where can I see Replicate incident history?
Replicate's official status page at status.replicate.com keeps a history of past incidents and maintenance, with timestamps and resolution notes. That is the canonical record for postmortems and recurring patterns. For broader context on how Replicate compares to other inference and media providers during the same window, this page and the wider TensorFeed status board track the ecosystem side by side.