Full Steps: llama.cpp on Windows with Intel Arc B580

Step 1 — Check your Windows Intel driver version

Open PowerShell:

Get-WmiObject Win32_VideoController | Select Name, DriverVersion

You want driver 31.0.101.6314 or newer for B580. If older, update from intel.com/arc-drivers before continuing.

Step 2 — Download pre-built llama.cpp (no compilation needed)

Go to https://github.com/ggml-org/llama.cpp/releases/latest and download:

llama-bXXXX-bin-win-vulkan-x64.zip

Or in PowerShell:

# Create a working directory
mkdir C:\llama.cpp
cd C:\llama.cpp

# Download latest release (Vulkan build)
$release = Invoke-RestMethod "<https://api.github.com/repos/ggml-org/llama.cpp/releases/latest>"
$asset = $release.assets | Where-Object { $_.name -like "*win-vulkan-x64*" }
Invoke-WebRequest $asset.browser_download_url -OutFile "llama-vulkan.zip"

# Extract
Expand-Archive llama-vulkan.zip -DestinationPath .

Step 3 — Copy your model from WSL2 to Windows

In PowerShell:

# WSL2 files are accessible at \\wsl$\Ubuntu\home\taoch\llama.cpp\
mkdir C:\llama.cpp\models
Copy-Item "\\wsl$\Ubuntu\home\taoch\llama.cpp\Qwen2.5-7B-Instruct-Q4_K_M.gguf" C:\llama.cpp\models\

Or from WSL2 terminal:

cp ~/llama.cpp/Qwen2.5-7B-Instruct-Q4_K_M.gguf /mnt/c/llama.cpp/models/

Step 4 — Verify both B580s are visible

In PowerShell: