Status
Is the API answering?
This page shows what we measure: whether the API answers, how each model did in the last hour of real requests, and our nightly timing run.
The API is answering
Our web server asked the API for this page, and the API answered. You can call 14 models now.
Models in the last hour
From real requests in the last 60 minutes. "Answered" is the share of requests that got an answer. "First byte" is the median time to the first part of a streamed answer.
14 with no recent requests.
Nightly timing
Each night we send one real request to each model and record how fast it answers. 0 of 14 models gave a timing in the last run. Measured 2026-10-30 03:00 UTC.
No timing in this run
These models gave no timing in the last run. A model can use the whole budget of our timing request before it writes an answer. A missed timing does not mean that the model is down.
How we choose the word
Operational: 95% or more of the requests in the last hour got an answer, and the median first byte is 10 seconds or less. Degraded: fewer than 95% got an answer, or the median first byte is more than 10 seconds. Down: fewer than half got an answer, with three or more requests that did not. A model with no requests in the hour shows "No recent requests".
This page shows no uptime over days and no incident history. We do not measure them, so we do not publish them.
If the API fails for you and this page looks normal, tell us.