Model status
Measured uptime, latency and model authenticity for every chat model, probed hourly from the public endpoint.
Models measured 0
How to read this table
There is no overall score, deliberately. A single number would have to fold availability, model identity and latency into one figure, and those pull in different directions: an 8.9s response is irrelevant to a batch job and fatal to a chat UI. Read the column you care about and sort by it.
- Excellent
- every probe in the last 24h answered
- Good
- at least 95% of probes answered
- Unstable
- fewer than 95% answered
- Down
- no probe in the last 24h answered
- Measuring
- fewer than 6 probes so far — too few to call
A model can sit behind several channels and the station gives no way to address one, so a bad channel shows up here statistically - as a mismatch rate or a latency tail, not as a named channel.