PyTorch 2.14 with CUDA 13.2
For PyTorch 2.14 with CUDA 13.2, pytorch release is 2.14; cuda build is CUDA 13.2; python versions is >=3.10, <=(3.15, 3.15t experimental); cudnn version is 9.24.0.43; gpu architectures compiled in, linux wheel is Turing(7.5), Ampere(8.0, 8.6), Hopper(9.0), Blackwell(10.0, 12.0+PTX), recorded from its source on 2026-09-14.
- Pairing
- PyTorch 2.14 with CUDA 13.2 our reading
- PyTorch release
- 2.14 verified
- CUDA build
- CUDA 13.2 verified
- Python versions
- >=3.10, <=(3.15, 3.15t experimental) verified
- cuDNN version
- 9.24.0.43 verified
- GPU architectures compiled in, Linux wheel
- Turing(7.5), Ampere(8.0, 8.6), Hopper(9.0), Blackwell(10.0, 12.0+PTX) verified
- Minimum driver (minor version compatibility)
- >= 580 verified
- Wheel index URL
- https://download.pytorch.org/whl/cu132 our reading
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
| 2.14 | >=3.10, <=(3.15, 3.15t experimental) | C++20 | CUDA 12.6 (CUDNN 9.10.2.21) (NCCL 2.29.3), CUDA 13.0 (CUDNN 9.24.0.43) (NCCL 2.30.7), CUDA 13.2 (CUDNN 9.24.0.43) (NCCL 2.30.7) | -- | ROCm 7.14 |
— raw.githubusercontent.com, retrieved 2026-09-14
Where each value comes from
This source states these in separate places, so each value is shown with the passage that states it.
GPU architectures compiled in, Linux wheel
| 13.2.1 | Turing(7.5), Ampere(8.0, 8.6), Hopper(9.0), Blackwell(10.0, 12.0+PTX) | +PTX available on linux builds only |
Minimum driver (minor version compatibility)
13.x | >= 580 | N/A
— all from raw.githubusercontent.com, retrieved 2026-09-14
Sources
- raw.githubusercontent.comhttps://raw.githubusercontent.com/pytorch/pytorch/main/RELEASE.md
- docs.nvidia.comhttps://docs.nvidia.com/cuda/cuda-toolkit-release-notes/index.html