Nightly packages for Python 3.14 or CUDA 13.x available for testing
We added a new nightly python package feed for CUDA 13.x. See Installation section below.
Python 3.14
Nightly packages now support python 3.14:
| Device | OS | cp314 | cp314t |
|---|---|---|---|
| CUDA 13.x | Linux x64 | Available | Available |
| CUDA 13.x | Windows | Available | None |
| CUDA 12.x | Linux x64 | Available | Available |
| CUDA 12.x | Windows | Available | None |
| CPU | Linux x64 | Available | Available |
| CPU | Linux ARM64 | Available | Available |
| CPU | Windows | Available | None |
| Tag | CPython Version & Build | Key Feature |
|---|---|---|
| cp314 | CPython 3.14 (Standard) | Includes the Global Interpreter Lock (GIL). |
| cp314t | CPython 3.14 (Free-Threaded) | Built without the GIL (or with a free-threading alternative). |
Installation
First, you might need uninstall existed packages like
pip uninstall onnxruntime onnxruntime-gpu -y
Then choose only one of the following based on cuda version (or CPU only) to install:
CUDA 13.x
pip install flatbuffers numpy packaging protobuf sympy coloredlogs
pip install --pre --index-url https://aiinfra.pkgs.visualstudio.com/PublicPackages/_packaging/ort-cuda-13-nightly/pypi/simple/ onnxruntime-gpu --no-deps
CUDA 12.x
pip install flatbuffers numpy packaging protobuf sympy coloredlogs
pip install --pre --index-url https://aiinfra.pkgs.visualstudio.com/PublicPackages/_packaging/ORT-Nightly/pypi/simple/ onnxruntime-gpu --no-deps
CPU Only
pip install flatbuffers numpy packaging protobuf sympy coloredlogs
pip install --pre --index-url https://aiinfra.pkgs.visualstudio.com/PublicPackages/_packaging/ORT-Nightly/pypi/simple/ onnxruntime --no-deps
If you encountered issues, please reply to this issue, or create a new issue and reference this for tracking.
Related issues: https://github.com/microsoft/onnxruntime/issues/26388 https://github.com/microsoft/onnxruntime/issues/26473 https://github.com/microsoft/onnxruntime/issues/26238
Thank you! It's asking for a username/password on my end though: https://github.com/microsoft/onnxruntime/issues/26568
onnxruntime requires coloredlogs (really necessary?), which causes dependency issues on Windows/py-314:
- Installs pyreadline (humanfriendly>=9.1->coloredlogs->onnxruntime), which is abandonware
- Python console raises AttributeError: module 'collections' has no attribute 'Callable'
Thanks a lot for that!
Are those good to use with python 3.14 FT? I see there are 3.14t wheels but when loading it on a FT enviroment I get the :
Python 3.14.0 free-threading build (main, Oct 7 2025, 15:35:12) [Clang 20.1.4 ] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import onnxruntime
<frozen importlib._bootstrap>:491: RuntimeWarning: The global interpreter lock (GIL) has been enabled to load module 'onnxruntime.capi.onnxruntime_pybind11_state', which has not declared that it can run safely without the GIL. To override this behavior and keep the GIL disabled (at your own risk), run with PYTHON_GIL=0 or -Xgil=0.
(used this wheel: onnxruntime-1.24.0.dev20251206001-cp314-cp314t-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl)
Is there an estimate as to when support for 3.14 will be available as an official release? We'd prefer not to have to special case Python 3.14 to use a custom nightly package.
Is there an estimate as to when support for 3.14 will be available as an official release? We'd prefer not to have to special case Python 3.14 to use a custom nightly package.
It will be next release 1.24.0 in Jan 2026 according to https://onnxruntime.ai/roadmap.
where i can get onnxruntime.so after running
pip install flatbuffers numpy packaging protobuf sympy coloredlogs
pip install --pre --index-url https://aiinfra.pkgs.visualstudio.com/PublicPackages/_packaging/ort-cuda-13-nightly/pypi/simple/ onnxruntime-gpu --no-deps
?
I dont see it in site-packages/onnxruntime or site-packages/onnxruntime-gpu
where i can get
onnxruntime.soafter running ? I dont see it insite-packages/onnxruntimeorsite-packages/onnxruntime-gpu
$ ls $CONDA_PREFIX/lib/python3.12/site-packages/onnxruntime/capi
__init__.py _ld_preload.py build_and_package_info.py libonnxruntime.so.1.24.0 libonnxruntime_providers_shared.so onnxruntime_collect_build_info.py onnxruntime_pybind11_state.cpython-312-x86_64-linux-gnu.so
__pycache__ _pybind_state.py convert_npz_to_onnx_adapter.py libonnxruntime_providers_cuda.so libonnxruntime_providers_tensorrt.so onnxruntime_inference_collection.py onnxruntime_validation.py
Hi,
I wonder what's the status of onnxruntime on macOS + python 3.14? Will it stop working, or can we build on our own?
@tianleiwu The pip file is missing in the servers, it returns this error:
(.venv) $ pip install --pre --index-url https://aiinfra.pkgs.visualstudio.com/PublicPackages/_packaging/ort-cuda-13-nightly/pypi/simple/ onnxruntime-gpu --no-deps
Looking in indexes: https://aiinfra.pkgs.visualstudio.com/PublicPackages/_packaging/ort-cuda-13-nightly/pypi/simple/
ERROR: Could not find a version that satisfies the requirement onnxruntime-gpu (from versions: none)
ERROR: No matching distribution found for onnxruntime-gpu
Could someone from the dev team clarify the open questions in this issue please? I'm about to include onnxruntime in my Python project, but I need support for Python 3.14 on Windows, Linux, and macOS (the table does not list macOS at all). I don't care for 3.14t, am I correct in assuming that this will not further delay the v1.24 release?
mac can not install and not found for python3.14. please release !!!!!!!!!
(torch4s)@MullerdeMacBook-Pro code % pip install --pre --index-url https://aiinfra.pkgs.visualstudio.com/PublicPackages/_packaging/ort-cuda-13-nightly/pypi/simple/ onnxruntime --no-deps
Looking in indexes: https://aiinfra.pkgs.visualstudio.com/PublicPackages/_packaging/ort-cuda-13-nightly/pypi/simple/ ERROR: Could not find a version that satisfies the requirement onnxruntime (from versions: none) ERROR: No matching distribution found for onnxruntime