PyTorch 2.6 with CUDA 11.8

For PyTorch 2.6 with CUDA 11.8, pytorch release is 2.6; cuda build is CUDA 11.8; python versions is >=3.9, <=3.13, (3.13t experimental); build script: base architecture list is TORCH_CUDA_ARCH_LIST="5.0;6.0;7.0;7.5;8.0;8.6"; build script: this cuda build's architecture line is TORCH_CUDA_ARCH_LIST="${TORCH_CUDA_ARCH_LIST};3.7;9.0", recorded from its source on 2026-09-16.

Values marked our reading are our classification of what the source says — the source does not print them in those words. The quote below is the evidence for each one; judge it yourself.

What the source says

| 2.6 | >=3.9, <=3.13, (3.13t experimental) | C++17 | CUDA 11.8, CUDA 12.4 (CUDNN 9.1.0.70) | CUDA 12.6 (CUDNN 9.5.1.17) | ROCm 6.2.4 |

— raw.githubusercontent.com, retrieved 2026-09-16

Where each value comes from

This source states these in separate places, so each value is shown with the passage that states it.

— all from raw.githubusercontent.com, retrieved 2026-09-16

Sources

• raw.githubusercontent.comhttps://raw.githubusercontent.com/pytorch/pytorch/main/RELEASE.md
• raw.githubusercontent.comhttps://raw.githubusercontent.com/pytorch/pytorch/release/2.6/.ci/manywheel/build_cuda.sh
• raw.githubusercontent.comhttps://raw.githubusercontent.com/pytorch/pytorch/release/2.6/.ci/pytorch/windows/cuda118.bat
• docs.nvidia.comhttps://docs.nvidia.com/cuda/cuda-toolkit-release-notes/index.html

Last verified against source: 2026-09-16. Due for re-check by 2026-12-15. This page as Markdown · OKF bundle · full dataset as JSON.