How to Check Your CUDA Version on Windows, Mac, and Linux

Find your CUDA version in three commands or less

CUDA is a toolkit that lets your graphics card (GPU) do computing work instead of just displaying images. If you use machine learning software, video editing tools, or scientific programs, they often need a specific CUDA version to run. The fastest way to check what you have installed is to open a terminal or command prompt and run a single command — the method depends on whether you use Windows, Mac, or Linux.

Most of the time, you will get an answer in under 10 seconds. If CUDA is not installed, the command will fail, and that tells you something too. This guide covers the exact commands to type and what the output means.

Key Takeaways

On Windows, open Command Prompt and type nvidia-smi to see your CUDA version in the top right corner of the output.
On Mac and Linux, open a terminal and type the same command: nvidia-smi, or try nvcc --version if that does not work.
If the command does not run, CUDA is either not installed or not in your system path, and you will need to install it or add it to your path.
The CUDA version shown by nvidia-smi is the driver version; the version shown by nvcc --version is the toolkit version — they can differ.
Some programs need a specific CUDA version, so write down the number you see and check your software's documentation before updating.

Check CUDA on Windows using Command Prompt

Open Command Prompt by pressing Windows key + R, typing cmd, and pressing Enter. At the prompt, type the following and press Enter:

nvidia-smi

If CUDA is installed and your graphics card is recognized, you will see a table with information about your GPU. Look at the top right corner of the output — you will see a line that says CUDA Version: X.X (for example, CUDA Version: 12.1). That number is your CUDA version. If you see nothing about CUDA, scroll up in the Command Prompt window to make sure you are reading the whole output.

If the command does not run and you see an error like "nvidia-smi is not recognized", CUDA is not installed, or the NVIDIA driver is not in your system path. You can try the second method below, or you may need to install CUDA from NVIDIA's website.

Check CUDA on Mac and Linux using the terminal

Open a terminal window. On Mac, press Command + Space, type terminal, and press Enter. On Linux, use your application menu or press Ctrl + Alt + T depending on your desktop environment.

Type the following command and press Enter:

nvidia-smi

The output is the same as on Windows — look for CUDA Version in the top right. If this command does not work on Mac, CUDA may not be installed, or you may be using an Apple Silicon Mac (M1, M2, M3), which does not support NVIDIA CUDA. Apple Silicon Macs use a different framework called Metal Performance Shaders instead.

If nvidia-smi does not work on Linux, try this command instead:

nvcc --version

This shows the CUDA toolkit version, which is sometimes different from the driver version shown by nvidia-smi. The toolkit version is what most software cares about, so if you see a version here, write it down.

Understand the difference between driver version and toolkit version

CUDA has two parts: the driver (which talks to your graphics card) and the toolkit (which contains the programming tools and libraries). The nvidia-smi command shows the driver version, and nvcc --version shows the toolkit version. They do not always match, and that is normal.

Most software cares about the toolkit version, not the driver version. If a program says it needs CUDA 11.8, check the output of nvcc --version first. If nvcc does not exist on your system, the toolkit is not installed, even if the driver is. You can install the toolkit separately from the driver.

The driver version is usually higher than the toolkit version. For example, you might have a driver version of 12.1 but a toolkit version of 11.8. This is fine — the driver is backward compatible, meaning newer drivers work with older toolkits.

What to do if CUDA is not installed

If both nvidia-smi and nvcc --version fail to run, CUDA is not installed on your computer. Before you install it, check whether your graphics card is supported. NVIDIA CUDA only works with NVIDIA graphics cards — if you have an AMD or Intel GPU, you will need a different toolkit.

To check your graphics card, open Device Manager on Windows (press Windows key + X and select Device Manager), expand Display adapters, and look for NVIDIA. On Mac, click the Apple menu, select About This Mac, and look under Graphics. On Linux, open a terminal and type lspci | grep -i nvidia.

If you have an NVIDIA card and want to install CUDA, visit NVIDIA's CUDA download page, select your operating system and graphics card, and follow their installation steps. The process takes 10 to 30 minutes depending on your internet speed and computer.

Add CUDA to your system path if commands do not work

Sometimes CUDA is installed but the commands do not run because the installation folder is not in your system path. The system path is a list of folders where your computer looks for programs to run.

On Windows, right-click This PC or My Computer, select Properties, click Advanced system settings, then Environment Variables. Under System variables, find the one called Path, click Edit, and add the CUDA bin folder (usually C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.1\bin, but the version number may differ). Click OK, close all windows, and restart Command Prompt. Then try nvidia-smi again.

On Mac and Linux, open a terminal and type:

export PATH=/usr/local/cuda/bin:$PATH

Then try nvcc --version. If this works, you can make the change permanent by adding that line to your shell configuration file (usually ~/.bashrc or ~/.zshrc on Mac).

Verify your CUDA version matches your software requirements

Once you know your CUDA version, check whether it matches what your software needs. Open the documentation or system requirements for the program you want to use and look for a CUDA version number. If your version is lower, you may need to update CUDA. If your version is higher, it usually still works because newer CUDA versions are backward compatible with older software.

Write down both your driver version (from nvidia-smi) and your toolkit version (from nvcc --version) before you make any changes. If updating CUDA breaks something, you will want to know what you started with. Some software is picky about exact versions, so do not assume a newer version will work without checking first.

Frequently Asked Questions

What if nvidia-smi shows a different CUDA version than nvcc?

nvidia-smi shows the driver version, and nvcc shows the toolkit version. They can differ, and that is normal. Most software cares about the toolkit version (from nvcc), so use that number when checking your software's requirements. The driver version is usually higher and does not need to match exactly.

Can I have multiple CUDA versions installed at the same time?

Yes. You can install multiple CUDA toolkit versions on the same computer, and software will use whichever one it is configured for. However, you can only have one driver installed. If you need multiple toolkits, install them in different folders and point each program to the one it needs.

Why does my Mac not show CUDA even though I installed it?

If you have an Apple Silicon Mac (M1, M2, M3, or newer), NVIDIA CUDA does not work on it. Apple Silicon uses a different architecture, and NVIDIA does not support it. You will need to use Metal Performance Shaders or another GPU framework instead. If you have an Intel Mac, make sure you installed the Mac version of CUDA, not the Linux or Windows version.

Do I need to update CUDA if my software works fine?

No. If your software runs without errors, your CUDA version is compatible and you do not need to update. Updating CUDA can sometimes break existing software, so only update if a program specifically requires a newer version or if you are installing new software that needs it.

What does it mean if nvidia-smi runs but shows no GPU?

This usually means your NVIDIA driver is installed but your graphics card is not being recognized by the system. Check Device Manager (Windows) or System Report (Mac) to confirm your GPU is listed. If it is listed but nvidia-smi does not see it, try restarting your computer or reinstalling the NVIDIA driver.

This guide is for general information only and is not medical, financial, legal, or other professional advice. For decisions specific to your situation, consult a qualified professional. See our Editorial Policy.