API gateway
TLS termination, key auth, routing to qwen-flash-next on SGLang, FP8.
90 days agotoday
No known incidents.
Performance · last 7 days
First token · p50
1.78s
Median over streaming requests, server-side.
First token · p95
8.13s
The tail is long-context prefill.
Output throughput
94tok/s
Per stream. An Unlimited key runs up to 5 concurrent streams.
Error rate · 7 days
2.2%
Includes client timeouts and
429.