Status
Is it working?
Two things we can state from measurements rather than from memory: whether the API answers, and which models responded in the most recent sweep.
The API is answering
This page is served by asking the API directly, so if you are reading this it answered. 45 models are available to call.
Models
We time every model with a real request once a day and publish the result on the models page. 36 of 38 produced a timing in the last run. Measured 2026-08-01 03:00 UTC.
No timing this run
These answer requests normally. They think for a variable length of time before replying, and when that runs long they exceed the fixed budget our timing run allows, so it records no measurement. They stay callable and their published speed is the last figure we did measure.
What this does not tell you
There is no uptime percentage, no latency percentile and no incident history here, because we do not measure them and a number we invented would make this page worth less than nothing. The daily timing run is a speed measurement, not an availability check: a model can stop answering between runs and still appear here as measured.
If the API is failing for you and this page looks healthy, tell us. We post incidents to the address on your account while they are happening rather than after they are resolved. Contact.