diff --git a/README.md b/README.md
index ee0c3dc..2ea16c9 100644
--- a/README.md
+++ b/README.md
@@ -68,6 +68,8 @@ export LD_LIBRARY_PATH=/usr/local/cuda/lib64:$LD_LIBRARY_PATH
 
 On Apple Silicon (e.g. Mac Mini M4), Metal is enabled by default and runs both the transformer and vision tower on the GPU. See [docs/backend/metal.md](docs/backend/metal.md) for building `vla.cpp` on macOS.
 
+On Windows, build inside WSL2 (Ubuntu). See [docs/backend/windows-wsl.md](docs/backend/windows-wsl.md) for the WSL2 + CUDA setup, including the GPU/toolkit prerequisites.
+
 ## Install simulators
 
 The eval scaffold under [`eval/`](eval/) supports two simulators end-to-end. Each setup script bootstraps an isolated Python 3.10 `uv` venv next to itself and clones the upstream sim repo. Both require [`uv`](https://github.com/astral-sh/uv) on `PATH`.
diff --git a/docs/backend/windows-wsl.md b/docs/backend/windows-wsl.md
new file mode 100644
index 0000000..68350b4
--- /dev/null
+++ b/docs/backend/windows-wsl.md
@@ -0,0 +1,117 @@
+# `vla.cpp` on Windows (WSL2 + CUDA)
+
+`vla.cpp` targets Linux and macOS. On Windows the supported path is **WSL2**
+with an Ubuntu distribution: the toolchain (`libzmq`, `protobuf`, `pkg-config`,
+the bash `patches/patch.sh` script) and the CUDA build all run natively inside
+the Linux environment, while still using the host NVIDIA GPU through the
+WSL CUDA driver.
+
+> **Status: draft.** Verified through toolchain + dependency setup on Ubuntu
+> 24.04 (WSL2) with the CUDA 12.6 toolkit and an RTX 4050 Laptop GPU
+> (`sm_89`). The end-to-end build and the SmolVLA run are being validated;
+> the [Results](#results) section is filled in once that completes.
+
+## Prerequisites
+
+A WSL2 distribution (this guide uses **Ubuntu 24.04**) and a recent **NVIDIA
+Windows driver** that exposes the GPU to WSL. Confirm the GPU is visible from
+inside WSL before building:
+
+```bash
+nvidia-smi --query-gpu=name,memory.total,driver_version --format=csv
+# e.g. NVIDIA GeForce RTX 4050 Laptop GPU, 6141 MiB, 556.29
+```
+
+> Note: the Windows driver ships the WSL CUDA *driver*, but not the CUDA
+> *toolkit* (`nvcc`). The toolkit is installed inside WSL, below.
+
+Install the build dependencies and a CUDA toolkit that matches your driver
+(driver 556.29 supports CUDA ≤ 12.6; CUDA ≥ 12.4 is required for the GCC 13
+shipped on Ubuntu 24.04):
+
+```bash
+sudo apt-get update
+sudo apt-get install -y libzmq3-dev libprotobuf-dev protobuf-compiler \
+    pkg-config build-essential cmake git wget
+
+# CUDA toolkit (nvcc) via the WSL-specific repo
+cd /tmp
+wget https://developer.download.nvidia.com/compute/cuda/repos/wsl-ubuntu/x86_64/cuda-keyring_1.1-1_all.deb
+sudo dpkg -i cuda-keyring_1.1-1_all.deb
+sudo apt-get update
+sudo apt-get install -y cuda-toolkit-12-6
+
+export PATH=/usr/local/cuda-12.6/bin:$PATH
+export LD_LIBRARY_PATH=/usr/local/cuda-12.6/lib64:$LD_LIBRARY_PATH
+nvcc --version   # confirm the toolkit is on PATH
+```
+
+### Windows `PATH` translation errors
+
+WSL imports the Windows `PATH` into the Linux environment. A stale Windows
+`PATH` entry that points at a drive no longer present produces, on every
+command:
+
+```
+<3>WSL (NNN) ERROR: UtilTranslatePathList:2852: Failed to translate E:\Some\Dir
+```
+
+It is harmless, but to remove it delete the dead entry from the Windows user
+`PATH` (PowerShell), then restart WSL:
+
+```powershell
+$p = [Environment]::GetEnvironmentVariable('Path','User') -split ';'
+$p = $p | Where-Object { $_ -and $_ -ne 'E:\Some\Dir' }
+[Environment]::SetEnvironmentVariable('Path', ($p -join ';'), 'User')
+```
+
+```powershell
+wsl --shutdown   # reopen WSL afterwards so the new PATH is picked up
+```
+
+## Configure & build
+
+```bash
+# Fetch llama.cpp at the pinned tag and apply the local patch
+bash patches/patch.sh
+
+# CUDA build. Set CMAKE_CUDA_ARCHITECTURES for your GPU (see the table in the
+# top-level README; RTX 40-series / Ada = 89).
+cmake -B build \
+    -DGGML_CUDA=ON \
+    -DGGML_CUDA_GRAPHS=ON \
+    -DCMAKE_BUILD_TYPE=Release \
+    -DCMAKE_CUDA_ARCHITECTURES=89
+cmake --build build -j"$(nproc)"
+```
+
+## GPU offload
+
+The VLA core selects its compute backend at load time. With a CUDA build it
+picks the CUDA backend; confirm it from the SmolVLA startup banner:
+
+```
+vla: backend = CUDA (device 0)
+```
+
+If you instead see `vla: backend = CPU (4 threads)`, the build did not pick up
+CUDA — rebuild from a clean `build/` and check `GGML_CUDA` is `ON` in the CMake
+cache and that `nvcc` was on `PATH` at configure time.
+
+## Run SmolVLA
+
+SmolVLA ships a combined GGUF plus a separate `mmproj` vision tower (see
+[Models](../../README.md#models); GGUF published at
+[`vrfai/smolvla-libero-gguf`](https://huggingface.co/vrfai/smolvla-libero-gguf)).
+
+```bash
+./build/vla-server "$VLA_MMPROJ" "$VLA_GGUF"
+# vla-server: bound to tcp://*:5555. ready.
+```
+
+Drive it with the LIBERO client (`--arch smolvla`) as described in the
+top-level README.
+
+## Results
+
+_Pending end-to-end validation on WSL2 + RTX 4050._