Overview

prod · eu-central · last 24 hours

Requests 18.4M
+12.6%
p95 latency 412 ms
−38 ms
Error rate 0.42%
+0.06%
Cost /1k $0.019
−8.1%

Latency distribution

p50 / p95 / p99 across 6 services

  • p50
  • p95
  • p99

Token spend

by model family

$3.8k total
  • deepseek-v4 46%
  • kim-k3 28%
  • qwen-4 18%
  • other 8%

Slowest traces

Trace Service Model Tokens Latency Status
4f8e2c91 gateway deepseek-v4 18.4k 412 ms ok
a12b7e04 retriever kim-k3 9.2k 890 ms slow
c90e3f55 summariser deepseek-v4 31.0k 1240 ms slow
77bd1180 router qwen-4 2.1k 96 ms ok