gpulse is a real-time GPU monitoring tool for the terminal. It supports NVIDIA (via NVML), AMD (via ROCm), Intel (via Level Zero), and Apple Silicon (via Metal/IOKit) GPUs. Key features include predictive OOM (out-of-memory) detection using linear regression and spike analysis, 7 dashboard view modes (Grid, Detail, List, Predict, Compare, Topology, Fleet), 15 color themes, per-process GPU attribution, Prometheus metrics export, and fleet monitoring via SSH. gpulse is free for local monitoring. Install with: brew tap gpulseai/gpulse && brew install gpulse. Available at https://gpulse.ai.

v2.2.0 — OOM Prediction is here

Never lose a training run
to OOM.

Real-time GPU monitoring with predictive memory leak detection. Supports Apple Silicon, NVIDIA, AMD, and Intel — all from a single terminal command.

Download Free See Features
Install gpulse
brew tap gpulseai/gpulse && brew install gpulse
gpulse dashboard monitoring 8 NVIDIA GPUs in grid view with Monokai theme, showing real-time VRAM usage, GPU utilization, temperature, power draw, OOM leak alert, and per-process attribution
Grid view — 8 NVIDIA GPUs with real-time monitoring and OOM leak detection alert

FEATURES

Everything you need to monitor GPUs

Memory leaks, thermal throttling, OOM crashes — gpulse detects problems before they cost you hours of compute.

Beautiful TUI Dashboard

A btop-quality terminal interface with 7 purpose-built views, Monokai default theme, and vim-style navigation. See memory, utilization, temperature, power draw, and per-process attribution at a glance. Ships with 15 themes. Customize everything with gpulse config.

Predictive OOM Detection

Three algorithms — linear trend analysis, spike detection, and composite scoring — provide real-time risk scoring for every GPU. View results in the dedicated Predict view with per-GPU risk levels, growth rates, and estimated time-to-OOM before your training run crashes.

Every GPU, One Tool

Supports Apple Silicon (Metal), NVIDIA (NVML), AMD (ROCm), and Intel (Level Zero). One binary, universal vendor coverage. Fleet monitoring for multi-machine setups coming soon.

SCALE

From laptop to fleet

Start with local monitoring. Scale to 20 machines with Pro. Go unlimited with Enterprise cloud integration.

Local

FREE

Monitor GPUs on your machine

Install and run. No account needed. Full OOM prediction, 15 themes, 7 view modes, and Prometheus endpoint included.

PRO

Fleet

$29/mo

Monitor up to 20 machines via SSH

SSH into cloud VMs (AWS, Azure, GCP) or on-prem servers. Bastion/jump host support. Email, Slack, and Discord alerts. Advanced hardware metrics.

ENT

Cloud-Scale

CUSTOM

Auto-discover GPU instances across clouds

Cloud API integration for AWS, Azure, and GCP. K8s DaemonSet deployment. Agent-based push for unlimited machines. RBAC, SSO, and audit logs.

Compare all features

WHO USES GPULSE

Built for everyone who works with GPUs

From solo researchers to platform teams managing hundreds of GPUs.

ML Engineers

Stop losing training runs to OOM crashes. Predict memory exhaustion hours before it happens.

Uses: Predict mode, Detail view, Alerts

Data Scientists

Know if your GPU is actually being used — or if you forgot .to('cuda') again.

Uses: Grid view, Process table, Framework detection

MLOps / Platform Teams

Fleet-wide GPU visibility across your cluster. Know who's using what, and where capacity is wasted.

Uses: Grid view, Prometheus endpoint, Compare mode

Mac Studio / Apple Silicon Users

First-class unified memory monitoring. See GPU share, memory pressure, and swap — no other tool does this.

Uses: Detail view, Memory pressure, Predict mode

SREs / DevOps

Thermal alerts before your GPUs throttle. Detect hardware failures before they cascade into incidents.

Uses: Prometheus endpoint, Compare mode, Alerts

Game Developers

Profile VRAM budgets and thermal throttling during development. Cross-vendor, always-on, zero setup.

Uses: Detail view, Grid view, Thermal monitoring

VIEW MODES

Seven ways to see your GPUs

Switch between views with a single keystroke. Each mode is purpose-built for a different monitoring task.

gpulse Grid view with 8 NVIDIA GPUs showing VRAM usage bars, GPU utilization, temperature, power draw, OOM alert, and per-process attribution
Grid View — 8 GPUs at a glance with OOM leak alert
gpulse Detail view showing GPU 4 NVIDIA RTX 4090 in critical state with extended metrics, process table, and memory history
Detail View — Deep metrics for a single GPU
gpulse Topology view showing GPU interconnect diagram with PCIe/NVLink connections and PRO pill badges
Topology View — GPU interconnect diagram with NVLink
gpulse Fleet view showing local machine GPU fleet overview with health status indicators and Pro upgrade prompt
Fleet View — Multi-machine monitoring overview

GET STARTED

Up and running in 30 seconds

No configuration files. No dependencies. One command to install, one command to launch.

1

Install

One command via Homebrew. No dependencies.

Terminal
brew tap gpulseai/gpulse && brew install gpulse
2

Launch

First launch runs a quick 4-question setup wizard to configure your preferences.

Terminal
gpulse dashboard
3

Monitor

Navigate views with single keystrokes. Preferences saved to config.toml.

d Detail
P Predict
c Compare
t Theme

PRICING

Free for individuals. Pro for teams.

Download and start monitoring in seconds. No account required, no credit card needed.

Free

$0 forever
  • Local GPU monitoring
  • All vendors (NVIDIA, Apple, AMD, Intel)
  • 15 color themes
  • 7 view modes
  • OOM prediction & leak detection
  • Prometheus endpoint
Download Now

Pro

Coming Soon
$29 /month
  • Everything in Free
  • Fleet monitoring (SSH, up to 20 machines)
  • Email, Slack & Discord alerts
  • Web dashboard
  • Multi-machine support
  • Priority support
Join Waitlist

Enterprise

Coming Soon
Custom
  • Everything in Pro
  • Cloud API integration (AWS/Azure/GCP)
  • Agent-based fleet monitoring
  • RBAC & SSO
  • GPU scheduling & orchestration
  • Audit logs & SLA
Contact Us

FAQ

Frequently asked questions

What GPUs are supported?
gpulse supports NVIDIA (via NVML), Apple Silicon (via Metal/IOKit), AMD (via ROCm), and Intel (via Level Zero). All major GPU vendors are supported out of the box.
Does gpulse work on Linux?
gpulse currently supports macOS (both Apple Silicon and Intel). Linux support is coming soon in a future release.
How does OOM prediction work?
gpulse uses three algorithms — linear trend analysis, spike detection, and composite scoring — to analyze memory growth patterns in real-time. When a leak is detected, it calculates the estimated time to out-of-memory and alerts you.
What data does gpulse collect?
None. gpulse monitors your local GPU hardware and does not send any data anywhere. There is no telemetry, no analytics, and no phone-home. Your data stays on your machine.
Is gpulse open source?
The free tier is free to use but the source code is not publicly available. gpulse is distributed as a compiled binary.
How do I report bugs or request features?
Email us at info@gpulse.ai or use the feedback form on this website. We read every message.

Stop guessing. Start monitoring.

One command to install. Zero configuration. Free forever for local monitoring.

Install gpulse
brew tap gpulseai/gpulse && brew install gpulse
Download Free Contact Us