Live · refreshes every minute
FastInfra API status: partial outage
This page shows the live status of the FastInfra inference API: the overall state, gateway request volume and p50/p95 latency over the last 15 minutes, and for every backend its current state, latency, 24-hour uptime and sample count. Data refreshes every 30 seconds from the same health probes that drive automatic failover, and a machine-readable copy is available at /status.json for monitors.
What our own probes see, unedited: per-model response time right now, 24-hour uptime from one-minute samples, and
gateway latency across every customer. Machine-readable at /status.json.
Generated 09:22:40 UTC.
Inference backends
Each backend is probed every minute. "Latency" is the time to first token of a one-token completion where the backend supports it, otherwise the model-list round trip. Overflow routing means a degraded backend does not usually mean failed requests for you.
| Backend | State | Latency now | Median 24h | Uptime 24h | Last check |
|---|---|---|---|---|---|
| Free-tier models (FastInfra · DEFAULT) | down | 10,019 ms | - | 0.00% | 09:22:17 UTC |
| Qwen3.6-27B (FastInfra GPU · H200) | operational | 536 ms | 520 ms | 100.00% | 09:22:18 UTC |
| Qwen3.6-27B (FastInfra GPU · H200B) | down | 2,882 ms | - | 0.00% | 09:22:21 UTC |
| LTX-2.5 video generation | operational | 309 ms | 586 ms | 100.00% | 09:22:21 UTC |
| Whisper speech-to-text + Kokoro / Chatterbox TTS · PRIMARY | operational | 62 ms | 62 ms | 100.00% | 09:22:21 UTC |
| ↳ Whisper speech-to-text | operational | ||||
| ↳ Kokoro-82M text-to-speech (tts-1) | operational | ||||
| ↳ Chatterbox Multilingual text-to-speech (chatterbox) | degraded | queue 4519 / 16 · 4 of 4 workers | |||
Related
Frequently asked questions
How often does the status page update?
Every 30 seconds. The page is served with a short cache and the JSON endpoint at /status.json returns the same snapshot for monitors and dashboards.
What does a degraded backend mean for my requests?
Requests for models on a degraded backend are routed to another healthy route automatically where one exists; if no route is healthy, the request fails fast with an error you can retry.
Where are incidents and changes announced?
Behaviour changes, new endpoints and limit increases are listed on the changelog. Ongoing incidents are reflected here in the backend state and in the gateway error rate.
Is there a status API?
Yes. GET /status.json returns the overall state, gateway metrics and per-backend state, latency, uptime and per-model component readiness.
Last reviewed October 15, 2026 by FastInfra. Prices and counts on this page come from the live catalog.