Status

Live availability of the caft API, per model. Updated every minute.

OperationalUpdated 2026-10-21 17:29 UTC

Success rate: 99.89%
Avg throughput: 59 tok/s

Models

Last 24 hours, hourly. Models appear once they have served requests in that window.

›glm-5.3Operational100%2.4 s TTFT · 91 tok/s

Success rate100%

24h ago17:29 UTC

Time to first token2.4 s

24h ago17:29 UTC

Throughput91 tok/s

24h ago17:29 UTC

›glm-5.3-flashOperational99.89%8.9 s TTFT · 59 tok/s

Success rate99.89%

24h ago17:29 UTC

Time to first token8.9 s

24h ago17:29 UTC

Throughput59 tok/s

24h ago17:29 UTC

What the numbers mean

Success rate: The share of requests that completed without an error. A model below 90% is shown as degraded.
Time to first token (TTFT): How long a streamed response took to produce its first token, averaged over the hour. A model's row shows its latest hour.
Throughput: Output tokens per second, averaged over the hour.

These are the service's own measurements of real traffic, unedited. A request that fails before producing any output is not charged; see the Service Level page.