llm-watch

apollo · Qwen3.8-27B-UD-IQ4_XS.gguf

idle

Model

Context used

—

Generation

— tok/s

Prompt

— tok/s

Draft accepted

—

GPU

Utilisation

—

VRAM

—

Junction temp

—

Power

—

Host · live, last 3 min

CPU

—

Memory

—

GPU history

—

Power

—

Token usage · the last 24 hours

generated prompt processed prompt from cache

Throughput and cache over the same range

Generation

— tok/s avg

Prompt processing

— tok/s avg

Cache hit

—

Draft accepted

—

All time

llama.cpp log

loading…

Running on apollo