Models on the device
Inference on a phone that works with no signal.
We have no published ONNX build. Our driver app is native iOS and works with no signal, and ONNX Runtime also runs on iOS, which is where a model would sit.
Inference on a phone that works with no signal.
A FastAPI service that loads an ONNX model.
A trained model exported from your framework to ONNX.
The same model checked on each platform it ships to.
Illustration. A dashed outline marks where the technology would sit. We have no published build on it.
A minor release.
The next minor release.
A patch release.
Marked latest.
Source: ONNX releases on GitHub, checked 6 November 2026. The page shows day and month without a year. ONNX Runtime describes itself as a production-grade AI engine for training and inferencing.
A native iOS driver app that works with no signal, with a FastAPI back end behind it.

A Python 3.13 and FastAPI product whose AI step calls a Claude-compatible API.
Python
PyTorch
TensorFlow
iOS
Android
DockerThe ONNX releases page marks version 1.23.2 as latest, published on 6 October. Versions 1.23.1 and 1.23.0 came out on 29 September and 18 September.
No. We say so on the first call and show the nearest work, which is a native iOS driver app and Python back ends.
Its site lists Linux, Windows, Mac, iOS, Android and web browsers.
You do, from the first commit. Repositories and cloud accounts stay in your name.
Thirty minutes on your model, your devices and your signal. No sales pitch.