Ollama, LM Studio and ComfyUI all run natively on Windows, so most people never need Linux for local AI. But a lot of AI tools, scripts and Docker containers are written for Linux first. Windows Subsystem for Linux (WSL) lets you run them on the same PC you game on, with GPU acceleration, and without dual-booting.
Forge PC has not tested this setup for this guide. The steps come from Microsoft's WSL documentation, NVIDIA's CUDA on WSL user guide, AMD's ROCm documentation and Ollama's Linux install guide, all checked on 15 November 2026.
Do you actually need WSL?
Probably not if you just want to chat with a model or generate images. Our guide to running AI models with Ollama or LM Studio is the easier route. WSL is worth it when:
a project's instructions only cover Linux
you want to use Docker containers with GPU access
you are learning tools that you will later run on a Linux server
Step 1: Update your graphics driver on Windows
This is the step people get wrong. With WSL, the GPU driver lives on the Windows side. NVIDIA's guide says the Windows driver "is the only driver you need to install" and tells you not to install any Linux display driver inside WSL; the CUDA driver is made available inside WSL automatically.
So start by installing the latest Windows driver for your card, using our graphics driver guide.
Step 2: Install WSL
Microsoft's one-command install works on Windows 11 and on Windows 10 version 2004 (build 19041) or later:
Right-click PowerShell and choose Run as administrator.
Run wsl --install.
Restart the PC.
This enables the features WSL needs and installs Ubuntu by default. The first time you open Ubuntu from the Start menu, it decompresses its files and asks you to create a Linux username and password.
GPU acceleration needs WSL 2. NVIDIA's guide says WSL 1 doesn't support it, and new installs use WSL 2 by default. Run wsl -l -v in PowerShell to check the version of each distribution. Microsoft's CUDA page also says you need WSL kernel version 5.10.43.3 or later, and says to get the latest kernel by selecting Check for updates in Windows Update.
Step 3: Check the GPU is visible (NVIDIA)
Open Ubuntu and run nvidia-smi. If it lists your graphics card, CUDA is available inside WSL. NVIDIA says existing Linux CUDA applications can then run unmodified. You only need to install the CUDA Toolkit for Linux inside WSL if you want to compile CUDA programs yourself, and Docker with the NVIDIA Container Toolkit is supported for containers.
NVIDIA's guide also notes that CUDA Toolkit 13.x supports Turing (RTX 20-series) and later GPUs.
Step 3 for AMD: use ROCm on WSL
AMD's route is ROCm. Its documentation for ROCm 7.2.1 lists support for Radeon RX 9000-series and selected RX 7000-series graphics cards, plus Ryzen APUs, and says WSL support uses the open-source ROCDXG component with Adrenalin 26.2.2 and ROCm 7.2.1. Follow AMD's WSL installation page for your exact card, because support is model-specific. Our guide to running local AI on a Radeon GPU covers the Windows-native options if your card isn't on the list.
Step 4: Install a test app
Ollama is a quick way to prove the GPU works. Its Linux documentation gives a one-line install script:
curl -fsSL https://ollama.com/install.sh | sh
Ollama's docs say AMD GPU users doing a manual install need its additional ROCm package. Once installed, run a small model and use ollama ps in a second terminal to see whether it is running on the GPU.
Tips
Read Microsoft's WSL best-practices guide before you move large projects in. Microsoft says it covers good practices for file storage, GPU acceleration and editing code with VS Code.
Don't run the same model server on both sides at once. A native Windows Ollama and a WSL Ollama would be competing for the same graphics card's VRAM.
Remember the disk space. Linux distributions, Docker images and models all count against your SSD.
Our verdict
WSL 2 is the easiest way to run Linux-first AI tools on a Windows gaming PC. Update the graphics driver in Windows, run wsl --install, confirm the GPU with nvidia-smi (NVIDIA) or follow AMD's ROCm on WSL steps, and never install a Linux GPU driver inside WSL. If you only want a chatbot or image generator, stick with the native Windows apps. Forge PC has not tested this setup; steps are from Microsoft, NVIDIA, AMD and Ollama documentation.