System status
Pre-launchPer-model uptime
Availability is recorded per model, from the day that model started serving. The strip starts at that date and fills to the right; empty cells are days that have not happened yet, not days we chose to leave out.
Qwen3.8 27B
No data yetPlatform
everything a request passes through before it reaches a modelThe gateway can be reachable while a model is not, so the two are measured separately. A model marked operational above means the gateway answered and that model served.
Right now
service level · from the status feedIncident ledger
Pre-launch: no persisted status record is presented.
Planned maintenance is not an incident and is never counted as one. Those windows are listed to the right, and are drawn in blue on the strips above.
Planned maintenance
Pre-launch: no planned-maintenance record is presented.
Windows are announced in advance and scheduled outside peak hours where the work allows. The raw uptime figure above includes them; the second figure excludes them.
This page is built from request metadata — counts, timestamps, latency, routing. Inferway's self-hosted serving path does not persist prompt or completion content. This scope does not cover Cloudflare edge processing or customer-selected downstream systems.
Transparency hubEvery performance figure states the concurrency level, the prompt and output shape, and the timestamp of the run. Cells stay empty until all three can be stated.
Load reports