macos-smi 0.1.1

nvidia-smi equivalent for Apple Silicon — real-time GPU/CPU/RAM monitor

macos-smi

nvidia-smi equivalent for Apple Silicon — real-time GPU/CPU/RAM monitor for anyone running LLM inference (MLX, llama.cpp, Ollama, etc.) on a MacBook.

╔══════════════════════════════════════════════════════════════════════╗
║  macos-smi                              Sun Apr 12 10:15:30 2026    ║
║  Apple M4 Pro  ·  64 GB unified memory                              ║
╠═══════════════════════════════╦══════════════════════════════════════╣
║  GPU                          ║  Memory (Unified)                    ║
║  Utilization :  73 %          ║  Used    :  24.3 GB                  ║
║  Power       :  18.2 W        ║  Total   :  64.0 GB  (38 %)          ║
╠═══════════════════════════════╬══════════════════════════════════════╣
║  CPU                          ║                                      ║
║  Utilization :  42 %          ║                                      ║
║  Power       :   8.4 W        ║                                      ║
╚═══════════════════════════════╩══════════════════════════════════════╝
  Ctrl-C to quit  ·  1 Hz  ·  local powermetrics

Install

# Using uv (recommended)
uv tool install macos-smi

# Using pip
pip3 install macos-smi

# From source
git clone https://github.com/tianszz/macos-smi.git
cd macos-smi && uv tool install .

Usage

# CPU + RAM only (no sudo required)
macos-smi

# Full stats including GPU utilization and power draw
sudo macos-smi

# Single snapshot and exit (like nvidia-smi default)
sudo macos-smi --no-loop

# Faster refresh
sudo macos-smi --interval 500

Requirements

• macOS on Apple Silicon (M1 and later)
• sudo access for GPU stats via powermetrics (CPU + RAM work without sudo)

How it works

GPU utilization and power draw are read from Apple's powermetrics utility (requires root). CPU utilization and RAM are read via psutil (no sudo needed). The display updates at 1 Hz by default using Rich.