Pre-built wheels that erase Flash Attention 3 installation headaches — now with Windows support! 🎉
🚀 Update: Windows Wheels Now Available! We've successfully built Flash Attention 3 wheels for Windows (CUDA 12.8 only for now).
Upstream PR: Windows compatibility fixes submitted to Dao-AILab/flash-attention#2047
Note: Until the PR is merged, wheels are built from our fork: windreamer/flash-attention@fix_windows_fa3
Pick the line that matches your setup (change cu128 / torch280 if needed):
# CUDA 12.8 + PyTorch 2.8.0
pip install flash_attn_3 \
--find-links https://windreamer.github.io/flash-attention3-wheels/cu128_torch280Visit the GitHub Pages site and choose the link that matches:
- CUDA 13.0 →
cu130_torch... - CUDA 12.9 →
cu129_torch... - CUDA 12.8 →
cu128_torch... - CUDA 12.6 →
cu126_torch...
Each page shows the one-liner you need.
- Weekly, every Sunday at 22:00 UTC
- On demand, by triggering the workflow manually if you need a fresher build
Releases are tagged with the build date (2025.10.15) so you always know how fresh your wheel is.
The build scripts and index generator are Apache-2.0.