pykeio/ort

Health v2 · maintenance only

Adoption not part of the score

2476 stars · 263 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

ort is a safe, fast Rust binding for Microsoft's ONNX Runtime, providing hardware-accelerated machine learning inference and training for ONNX models. It supports many execution providers (CUDA, TensorRT, OpenVINO, CoreML, etc.) and alternative pure-Rust backends, and is production-ready.

Use cases

• run ONNX models in Rust with GPU acceleration
• deploy PyTorch or TensorFlow models in production from Rust
• fast on-device ML inference in a Rust application
• fine-tune or train ONNX models in Rust
• run transformer models like BERT or YOLOv8 efficiently
• compile ML inference to WebAssembly

When to choose

• you need high-performance ML inference in a Rust codebase
• you want hardware-accelerator support across NVIDIA, Intel, Qualcomm, and Apple
• you need to deploy models on-device or on servers with low latency
• you want a maintained, production-ready alternative to onnxruntime-rs

When to avoid

• your project is in Python and PyTorch directly is sufficient
• you need a pure-Rust ML stack with no C++ dependency (though alternative backends exist)
• you require the stable 1.x API rather than the 2.0 release-candidate

Facets

library · maturity active

machine-learning llm-inference gpu-computing sdk machine-learning deep-learning large-language-models developer-tools rust cross-platform wasm onnx onnxruntime inference training rust-binding hardware-acceleration execution-providers gpu

Member repositories

For agents

markdown · JSON · MCP: product_card(name="pykeio/ort")