ONNX Models

ONNX is the Open Neural Network Exchange, and we take that name to heart! Many members of the community upload their ONNX models to various repositories, and we want to make it easy for you to find them. Check out our featured Foundry Local Models collection, or explore the popular repositories below:

🏭

Foundry Local Models Featured!

Discover AI models optimized for local deployment on NPUs, GPUs, and CPUs. Search and filter by device support, family, and more.

ONNX Community (hosted on Hugging Face)

Access, share, and discuss ONNX models published to Hugging Face by members of the ONNX Community.

Model Zoo (hosted on ONNX website)

Explore a wide variety of pre-trained ONNX models curated by the ONNX community.

AMD Models (hosted on Hugging Face)

Various quark-quantized models are graciously provided by AMD, ready to run using the ONNX Runtime generate() api!

Qualcomm ONNX Models (hosted on Hugging Face)

Discover ONNX-compatible models optimized for Qualcomm hardware on Hugging Face.

All ONNX Models on Hugging Face

Access all ONNX models available on the Hugging Face Model Hub here!

Can't find what you're looking for?

Convert to ONNX, optimize, and quantize your own models quickly and easily with Olive. Here's a quick snippet showing you how easy it can be done:

pip install olive-ai onnxruntime optimum

olive auto-opt -m microsoft/Phi-3-mini-4k-instruct -o models/phi3-mini-4k