pyke
pyke is a verified GitHub organization that publishes open-source developer tools and libraries, with a focus on Rust and machine learning. Its flagship project, ort, is a Rust gateway to ONNX Runtime and other ONNX inference engines for efficient, hardware-accelerated model deployment.
At a Glance
- Rust developers
- Machine-learning engineers
- AI/ML infrastructure teams
- Open-source and research projects
- +3 more
AI Tools by pyke
(1)
ort
Rust ONNX Runtime Inference Library
Discussions
No discussions yet
Be the first to start a discussion about pyke
Latest News
ort v2.0.0-rc.14 released with ONNX Runtime 1.31.0 update, CUDA 13 support work, WebGPU fixes, performance improvements and training memory fixes.
ort v2.0.0-rc.13 released with ONNX Runtime 1.28, clearer execution-provider behavior, custom-operator ergonomics and improved diagnostics.
ort v2.0.0-rc.12 released with multiversioning, automatic device selection, CUDA 13 support and build attestations.
ort v2.0.0-rc.11 released with WebNN and WASM execution providers for ort-web and updates to ort-tract and ort-candle.
Products & Services
Open-source Rust interface for hardware-accelerated ONNX model inference and training, primarily wrapping Microsoft ONNX Runtime while also supporting compatible pure-Rust backends.
Low-level Rust bindings/system integration layer used to link and interact with ONNX Runtime binaries.
Web/WASM backend for running ONNX inference in browser and web applications, including WebNN and WASM execution providers.
Pure-Rust alternative ONNX inference backend based on tract.
Market Position
pyke’s flagship ort is positioned as the Rust gateway to ONNX Runtime and as a faster, lower-overhead and more ergonomic alternative to direct framework inference for production deployment. It differentiates from other ONNX Rust bindings through broad hardware execution-provider coverage, prebuilt binaries, web support, training APIs and alternative pure-Rust backends; comparable options include Microsoft’s ONNX Runtime Rust bindings, the older onnxruntime-rs crate, tract-based integrations and Candle-based runtimes.
Founding Story
The public record presents pyke as an open-source GitHub organization rather than a conventional venture-backed company. The organization account was created in March 2020; its ort repository says the library was based on the inactive onnxruntime-rs crate and aims to provide an ergonomic, fast Rust interface for deploying ONNX models.
Business Model
Revenue Model
The software is open source under Apache-2.0/MIT licensing and distributed through GitHub and crates.io. Support is funded through sponsorship/backer contributions via Open Collective; the project invites businesses to sponsor it in exchange for priority support and issue resolution.
Pricing Tiers
ort is available as a public Rust crate/repository under Apache-2.0/MIT licensing.
Sponsors support development and receive priority support and issue resolution, according to the official documentation.
Target Markets
- Rust developers
- Machine-learning engineers
- AI/ML infrastructure teams
- Open-source and research projects
- Edge/on-device application developers
- WebAssembly/browser developers
- Production deployment of PyTorch, TensorFlow, Keras, scikit-learn and PaddlePaddle models after ONNX conversion
- Low-latency inference on GPUs, NPUs and other accelerators
- On-device and edge ML applications
- Browser and WebAssembly ML inference
- Rust services and datacenter inference
- Local/private computer-vision, speech and embedding applications
- Wasmtime / Bytecode Alliance
- SurrealDB
- Supabase
- Google Magika