SiliconPulse Download
SiliconPulse app icon

See what your Mac
is really doing.

A lightweight, native monitor for Apple Silicon and the local AI models running on it. CPU, GPU and unified memory, your processes, and which model is loaded and generating, in one calm window and your menu bar.

Download for Mac View source

Free · macOS 14 or later · Apple Silicon and Intel · Release notes

The SiliconPulse Overview with a real oMLX model loaded: CPU, GPU and unified memory with live charts, the active model, disk, network and top processes

The whole Mac at a glance

CPU with performance and efficiency cores, GPU, and unified memory with pressure, compression and swap. Everything that matters on Apple Silicon, without the noise.

Your local models, first

Which model is loaded, in which runner, how much memory it takes and whether it is generating right now, with a live pulse while it works.

Lives in the menu bar

A live summary like “Qwen3 · GPU 91% · 38 tok/s” and a popover with the essentials. Under 1% CPU while it sits there.

Local AI

Is my model actually using the GPU?

SiliconPulse finds the runners on your Mac and asks them, on 127.0.0.1 only, what they are doing.

  • oMLX, Ollama, LM Studio, llama.cpp, MLX and any OpenAI-compatible server
  • Loaded model, model memory, context and GPU offload, as the runner reports them
  • The runner's own CPU, memory and GPU time, from its process tree
  • Tokens per second only when the runner really publishes it
  • System GPU is always the whole Mac, never credited to a model

Shown: a real oMLX server serving Qwen2.5 0.5B.

The AI Models screen with a real oMLX model: model name, tokens per second reported by oMLX, runner GPU time, model memory and runner CPU

Menu bar

The answer is one click away.

Health, CPU, GPU and memory, and the active model with what it is doing. Open the full window only when you want the details.

  • Choose what the menu bar shows, or just the icon
  • Starts quietly at login if you want it to
  • Slows its sampling down when nothing is on screen
The SiliconPulse menu bar popover with CPU, GPU, memory and the active oMLX model

Honest numbers

No invented values.

macOS does not expose everything. When something is measured, SiliconPulse shows it plainly; when it is inferred, it says so; when it is not available, it says why. Hover any value to see where it comes from.

Exact read directly from the kernel Reported by macOS or the runner Calculated from measured counters Estimated marked on screen Unavailable with the reason
Memory screen with used memory, pressure, swap, headroom and a composition bar
Memory: what is used, cached and free, and whether macOS is under pressure.
Processes screen: a native table with filters and search
Processes: a native table with filters, search and an inspector.

Private and light

Everything stays on your Mac.

0analytics, accounts or uploads
127.0.0.1the only address it ever contacts
<1%CPU while in the menu bar
1sampling loop, native Swift

No Electron, no web views, no telemetry. History lives in memory and is gone when you quit. The source is on GitHub.

Install

  1. Download SiliconPulse.dmg.
  2. Open it and drag SiliconPulse to Applications.
  3. Open SiliconPulse from Applications. It lives in your menu bar.

If macOS says it could not verify that SiliconPulse is free of malware, open System Settings › Privacy & Security and click Open Anyway. You only need to do this once; the release notes say when a build is notarized.

Download for Mac

macOS 14 or later · Universal app

Screenshots come from SiliconPulse's automated tests on a virtual Mac, which reports no GPU utilization; on a physical Apple Silicon Mac the GPU column is live. Model names marked “mock” are test stand-ins.