llm-host

…

Serving

—

Loading

Models

Statistics

Generation

—

Prompt processing

—

Time to first token

—

Requests

—

KV cache

—

GPU

—

VRAM

—

MTP acceptance

—

speculative tokens kept

Prefix cache hits

—

prompt tokens reused

Generation tok/s

10 min agonow

Prompt processing tok/s

10 min agonow

KV cache %

10 min agonow

GPU utilization %

10 min agonow

GPU temperature °C

10 min ago0–100 °Cnow

GPU power W

10 min agonow

Requests

Finished	Duration	TTFT	Prompt tok	Prompt sec	Prompt tok/s	Generated tok	Generated sec	Gen tok/s	Peak KV

Server log

…