mlxtop
TUI monitor showing memory, paging, GPU load and tokens/s for local LLM runtimes.
Tested on Linux
Use it
Linux x86_64
You need first
- Runtimes:
Rust>=1.88 for build from source
download install script for reading
curl -fsSL https://raw.githubusercontent.com/maximpri/mlxtop/main/scripts/install.sh -o /tmp/mlxtop-install.shinstall mlxtop via script
sh /tmp/mlxtop-install.sh
Check that it works
mlxtop --versionThe output should contain “mlxtop 3.1.2”.
First thing to try
mlxtop --onceRun mlxtop in a terminal to open the live monitor, or mlxtop --once for a one-time report.
What it does
- Check memory, paging and GPU bottlenecksTested
How to use it
- Run
mlxtop --onceto print a static report from a 1-second sample. - Read the memory, paging and GPU readings in the report.
- Run
- Get a plain verdict on what is slowing things downTested
How to use it
- Run
mlxtop --once. - Read the health verdict in the static report.
- Run
- See live generation and prefill speedNot shown working
How far pickfit got
Linux x86_64
- Installs
- Starts
- 2 of 3 features work
- Works offline
macOS and Windows: the project says it runs there; pickfit hasn’t tried it.
Good for
- find whether memory, paging or GPU slows a local model
- watch live generation and prefill speed with cache reuse
- check memory and GPU readings with no model loaded
- check runtime connection and available readings
- inspect model processes CPU, memory and page-ins
How pickfit tested it
Each run starts from a clean machine and follows the steps above. How pickfit verifies
Linux x86_64
Its main function works
- Replayed from scratch
- Works offline
- Checked on 30 Oct 2026
- Version 3.1.2