Service status
Current API health
This page checks the API health endpoint from your browser when it loads. It does not keep an uptime history.
Open the health endpoint · checked from your browser, not a monitoring history.
Prelaunch model-server p95 · single-GPU test
Two concurrent requests · one RTX PRO 6000
The local model-server test ran 1,000 public requests with both profiles resident: 700 s1-fast text requests and 300 s1-pro requests, including 60 image requests. The s1-pro mix was 4:1 text to image.
| Profile | Requests | p50 | p95 | Errors |
|---|---|---|---|---|
| s1-fast | 700 | 36 ms | 244 ms | 0 |
| s1-pro | 300 | 208 ms | 644 ms | 0 |
Source: authenticated launch-mix receipt, 27 September 2026. Model-server latency from a prelaunch GPU test; it excludes the API gateway and network time and is not production or end-to-end service latency, a service-level promise, or a latency guarantee. The s1-pro row was measured on the prelaunch Winnow-12B Q8 implementation; the EU tier now serves s1-pro on Surogate Rune 26B-A4B v3, and the current end-to-end service-latency measurement is published on the landing page. Open the published receipt summary.
Planned maintenance is announced by email to account holders.