Skip to content

How-ToGPU

How to Use NVIDIA Project G-Assist: Install, Requirements and Useful Commands

NVIDIA's Project G-Assist is a free AI assistant that runs locally on GeForce RTX GPUs with 6GB+ VRAM. Requirements, installing it from the NVIDIA app's Discover section, opening it with Alt+G, Reasoning vs Flash mode and useful commands.

By , Systems & Technical Editor

4 min read

How to Use NVIDIA Project G-Assist: Install, Requirements and Useful Commands

Project G-Assist is an AI assistant from NVIDIA that runs on your own GeForce RTX graphics card rather than in the cloud. You type or speak a request, such as "What's my GPU utilization?" or "Optimize my graphics for Cyberpunk 2077", and it answers or changes the setting. This guide is based on NVIDIA's official G-Assist page and release notes.

What G-Assist is (and isn't)

NVIDIA describes G-Assist as an assistant that "helps you tune, control, and optimize your system". It runs locally on your RTX GPU, so NVIDIA says it's free to use and can work offline. It's built on a small language model, and NVIDIA's release notes say it isn't intended to be a broad conversational AI like a cloud chatbot. NVIDIA still labels it "Project G-Assist" and describes it as experimental.

NVIDIA's earlier release notes say G-Assist uses a Llama-based instruct model with 8 billion parameters. They also warn that when you prompt it while gaming, the GPU briefly allocates some of its power to AI inference, so you may see a short dip in frame rate for a few seconds.

Requirements

NVIDIA lists these system requirements:

Requirement

NVIDIA's minimum

Operating system

Windows 10 or Windows 11

Graphics card

Any GeForce RTX 20, 30, 40 or 50 series GPU with at least 6GB of VRAM (desktop or laptop), or RTX PRO equivalent

Voice commands

GeForce RTX 30 series or newer

CPU

Intel Pentium G, Core i3/i5/i7 or higher; AMD FX, Ryzen 3/5/7/9, Threadripper or higher

Disk space

7GB for the System Assistant, plus 3GB for voice commands

Software

NVIDIA driver 580.97 or newer, NVIDIA App 11.0.7 or newer

Language

English

VRAM is the one to watch. NVIDIA says that if you use G-Assist while a game is running, you should have at least 6GB of free VRAM for Reasoning mode, or 4.5GB for Flash mode, on top of what the game uses. On cards with less VRAM, a demanding game may not leave that much free. Our guide to checking VRAM usage in Windows shows how to see what your games use.

Step 1: Update your driver and the NVIDIA app

Install the latest GeForce driver and NVIDIA app. Our guide to updating graphics drivers covers the safest way to do it.

Step 2: Install G-Assist from the NVIDIA app

NVIDIA says G-Assist is installed from the Discover section of the NVIDIA app. NVIDIA lists separate disk space for the System Assistant (7GB) and voice commands (3GB); voice commands need an RTX 30 series or newer card.

Step 3: Open G-Assist

Once it's installed, NVIDIA says you can open G-Assist from the NVIDIA app overlay or by pressing Alt+G. NVIDIA's latest release notes also mention a Reasoning On/Off button in the bottom bar: Reasoning is on by default for accuracy, and turning it off switches to the faster Flash mode.

Commands worth trying

NVIDIA publishes a list of supported functions. Useful ones include:

  • System info: "What's my GPU utilization?", "What's my CPU utilization?", "How much VRAM does my GPU have?" and "Which NVIDIA driver do I have installed?"

  • Power: "How much power is my GPU using?" reports the GPU's average power draw over the last 60 seconds.

  • Game settings: "Optimize my graphics for [game]" applies recommended settings, and "Undo that optimization" reverts them. NVIDIA notes the game may need closing or restarting first.

  • Frame rate limits: "Set a Frame Rate Limit to 100" (needs a running game) and "Turn off my frame rate limiter". Our guide to capping FPS on NVIDIA and AMD explains why you might want to.

  • Monitor: "Set my monitor to 144Hz" and "What's the highest refresh rate my monitor supports?"

  • Recording: "Save a shadowplay clip" and "Start recording my gameplay". See our guide to recording gameplay with the NVIDIA app for the full feature.

  • Graphs: "Graph my CPU and GPU usage" plots metrics over time.

  • Overclocking: G-Assist can overclock the GPU core in 15MHz steps up to 60MHz, and "Remove my GPU overclock" reverts it.

G-Assist also supports plug-ins. NVIDIA lists community plug-ins for brands including Logitech, Corsair, MSI, Discord, Twitch and Spotify, and some device controls (such as Corsair peripheral settings) need the maker's software installed.

Tips

  • If a command doesn't work, check NVIDIA's supported functions list; G-Assist is designed around specific commands rather than open-ended chat.

  • For deeper monitoring or manual tuning, dedicated tools still offer more control. Our guide to monitoring temps and FPS with HWiNFO and Afterburner covers the usual setup.

  • Send feedback from the NVIDIA app if something goes wrong; NVIDIA says feedback shapes what it adds next.

Our verdict

G-Assist is a free, local way to check your PC's stats and change NVIDIA settings by typing or speaking, and it's easy to install from the NVIDIA app's Discover section. You need an RTX card with at least 6GB of VRAM, and for use in games, enough spare VRAM on top. Treat it as a convenient shortcut for settings and diagnostics rather than a general-purpose chatbot.

The bottom line

G-Assist is a free, local way to check your PC's stats and change NVIDIA settings by typing or speaking, and it's easy to install from the NVIDIA app's Discover section. You need an RTX card with at least 6GB of VRAM, and for use in games, enough spare VRAM on top. Treat it as a convenient shortcut for settings and diagnostics rather than a general-purpose chatbot.

Sources

  1. 01NVIDIA, Project G-Assist (product page, release highlights, supported functions and system requirements)

About the author

Marcus Thorne

Systems & Technical Editor

Marcus writes Forge PC's step-by-step build, troubleshooting and system optimisation guides. His background spans Windows kernel and power-state diagnostics, UEFI/BIOS configuration, NVMe storage arrays, ATX 3.1 power delivery and running local AI and LLM workloads on consumer graphics cards.

More from Marcus Thorne

Read next

ComfyUI vs SwarmUI: Which Local AI Image Generator Should You Use?

Versus

ComfyUI vs SwarmUI: Which Local AI Image Generator Should You Use?

ComfyUI's node graph gives full control over local AI image, video and audio workflows; SwarmUI adds a simpler Generate tab and can run ComfyUI as its back end, with the graph a tab away. SwarmUI suits beginners, ComfyUI power users.

Callum Reid4 min