InferGauge
://live
-
-
0s
elapsed
0
users
0
req/s
connecting
−
0
+
Stop test
Performance score
-
no data
Response
-
avg
-
p95
-
p99
-
ttft p95
-
inter-token p95
Load
-
active / target
-
in flight
-
completed
-
error rate
-
goodput
-
quality (sampled)
Tokens
-
input
-
output
-
tokens/s now
-
out/request
Cost
-
spend
-
per request
-
projected monthly
-
peak ceiling · 24/7
Latency & load timeline
p95 latency
avg latency
users
SLA limit
event
Saturation curve · p95 vs users
each dot = 1s window · the knee is your ceiling
SLA validation
Insights & recommendations
Errors & retries