Independently measured availability of the NovAI gateway. We publish real probe data — including incidents — because unverifiable uptime claims are worthless.
No incidents recorded in the last 30 days (no window of 3+ consecutive failed probes).
An independent probe requests the public GET /v1/models endpoint every 5 minutes from the origin region and records HTTP status and latency. A probe counts as “ok” only on HTTP 200 with a valid model list. Uptime = ok probes / total probes in the window. Raw data: /status.json.
Last probe: 2026-10-12T16:10:01Z · probes in window: 298. Related: pricing & policy transparency · price leaderboard.