PyTorch 2.14.0
- Released
- 2026-09-02
- Official CUDA wheels
- Official ROCm wheels
- Supported Python versions
Which GPUs each CUDA build runs on
Each CUDA build of PyTorch 2.14.0 contains compiled code for a fixed list of GPU architectures. A card runs a build when the list includes its architecture (a build compiled for 8.6 also covers 8.9 cards like the RTX 4090, since newer minor versions of the same generation run older code). If the list doesn't cover your card, PyTorch installs fine but fails at runtime with no kernel image is available for execution on the device.
| CUDA build | Compiled for (compute capability) | GPU generations | Install |
|---|---|---|---|
| 12.6 | 5.0, 6.0, 7.0, 7.5, 8.0, 8.6, 9.0 | Maxwell, Pascal, Volta, Turing, Ampere, Ada Lovelace, Hopper | pip install torch==2.14.0 --index-url https://download.pytorch.org/whl/cu126 |
| 13.0 default on PyPI | 7.5, 8.0, 8.6, 9.0, 10.0, 12.0 | Turing, Ampere, Ada Lovelace, Hopper, Blackwell | pip install torch==2.14.0 |
| 13.2 | 7.5, 8.0, 8.6, 9.0, 10.0, 12.0 | Turing, Ampere, Ada Lovelace, Hopper, Blackwell | pip install torch==2.14.0 --index-url https://download.pytorch.org/whl/cu132 |
Heads-up for older cards: a plain pip install torch==2.14.0 gets the CUDA 13.0 build, which has no Maxwell, Pascal, or Volta code. On those GPUs, install the CUDA 12.6 build instead.
Linux x86_64 wheels. Your NVIDIA driver must also be new enough for the CUDA build — see its CUDA page.
Added 2026-09-27. Release date from PyPI upload metadata. Build variants re-verified 2026-09-27 against the files actually published: download.pytorch.org/whl/<variant>/torch/ (x86_64 Linux) for the CUDA and ROCm builds, and PyPI plus that index for Python versions. cudaArchs is each CUDA build's TORCH_CUDA_ARCH_LIST for x86_64 Linux wheels, taken by running the arch-selection code in PyTorch's own build scripts at this release (pytorch/builder release branch manywheel/build_cuda.sh up to 2.5, pytorch .ci/manywheel/build_cuda.sh for 2.6-2.12, .ci/manywheel/build_env_setup.py from 2.13); +PTX suffixes are omitted. defaultPipCuda: the CUDA version a plain `pip install torch` gets on Linux x86_64, from the PyPI wheel's own dependency metadata. windowsCudaVersions: the CUDA builds with win_amd64 wheels on the same index (2026-09-27); Windows wheels are assumed to use the Linux architecture lists, as PyTorch's RELEASE.md states for recent releases. rocmArchs: PYTORCH_ROCM_ARCH per ROCm build, from pytorch .ci/docker/manywheel/build.sh at this tag (manylinux2_28-builder:rocm* branch); the ROCm 7.14 build uses the TheRock-based image and drops gfx900/gfx906, the 7.2 legacy build keeps them. xpuBuild: an Intel GPU (XPU) build of this release is on download.pytorch.org/whl/xpu (checked 2026-09-27).