Library / SDK
Inference & Serving
ONNX Runtime
1Cross 2platform 3accelerated 4engine 5for 6ONNX
the six Ws · specification
W1
Who
Developed and maintained by Microsoft with an open source community.
W2
What
ONNX Runtime is a cross platform accelerated engine for running ONNX format machine learning models in production.
W3
Where
Hosted on GitHub at microsoft/onnxruntime and distributed via PyPI, NuGet and npm.
W4
When
Released in 2018 and actively maintained.
W5
Why
It lets teams deploy a single exported model across CPUs, GPUs, mobile and edge hardware with consistent performance.
W6
With
Supports execution providers for CUDA, TensorRT, OpenVINO, DirectML and CoreML among others.
W7
Watch
MIT licensed, open source, maintained by Microsoft, roughly 21k GitHub stars.
alternatives
works with