---
id: onnx-weekly
version: "1.23.0.dev20260805"
license: Apache-2.0
license_treatment: permissive
maintenance: active
---

# onnx-weekly — Open Neural Network Exchange

License: permissive · Maintenance: active · Downloads: 264.5K/mo

## What it is and what it does

onnx-weekly is a development release of the ONNX Python package, distributed weekly to enable early testing and experimentation with the latest ONNX specification and tooling. It provides APIs for loading, creating, validating, and manipulating ONNX computation graphs—the standardized intermediate representation used across deep learning and traditional ML frameworks.

The package serves as a bridge for model interoperability: you can export models from frameworks like PyTorch or TensorFlow into ONNX format, then load and run them in different environments or hardware. It includes utilities for shape and type inference, graph optimization, and opset version conversion. As a weekly build, it tracks the main development branch and is intended for developers who want to test new features or contribute to ONNX before stable releases.

Use it for:

- Export trained models from PyTorch, TensorFlow, or other frameworks to ONNX format for cross-platform deployment
- Load and inspect ONNX model graphs to understand model structure, operators, and data types
- Test new ONNX specification features and operators before they appear in stable releases
- Validate model compatibility and perform shape/type inference on computation graphs
- Convert models between different ONNX opset versions for compatibility with target inference engines

## Worth the install?

AI-flagged interpretation of the facts above — verify before relying.

onnx-weekly provides a Python package for working with ONNX (Open Neural Network Exchange) models—an open format for representing AI models with a computation graph, built-in operators, and standard data types for model interoperability and inference.

Yes, if you are actively developing with ONNX or need the latest features for testing and experimentation. The weekly release cadence, active maintenance, permissive Apache-2.0 license, and broad platform support make it a low-risk choice for early adopters. However, if you need production stability, use the stable onnx package instead. No known security vulnerabilities as of 2026-08-14.

## Install

pip install onnx-weekly
uv add onnx-weekly
poetry add onnx-weekly

## Installing onnx-weekly

Before you install: Medium install friction with prebuilt wheels across multiple Python versions (3.10–3.14) and platforms (macOS, Linux, Windows, WebAssembly). Active maintenance with a recent release; last commit 2026-08-14. Depends on numpy, protobuf, typing_extensions, and ml_dtypes.

License in practice: Apache-2.0 permissive license allows commercial and private use with minimal restrictions; you must include a copy of the license and state significant changes.

Quickstart:

pip install onnx-weekly

import onnx

# Load or create an ONNX model
model = onnx.load('model.onnx')

Requires Python ≥3.10; numpy, protobuf, typing_extensions, and ml_dtypes must be installed as runtime dependencies.

Verify before relying:

- Whether onnx-weekly is intended for production use or experimental/pre-release testing only
- Specific differences between onnx-weekly and the stable onnx package beyond version timing
- Performance characteristics or known limitations of the development release

## Package facts

- License: Apache-2.0 (permissive)
- Python support: supports_current
- Install friction: medium
- Maintenance: active
- Downloads: 264.5K/month (top 15,000 on PyPI)
- Known vulnerabilities: none known

## Tags

neural network model format, onnx model interchange, ai model serialization, deep learning model exchange, inference model framework, cross-framework model compatibility, model graph representation, model-interchange, neural-networks

[View on SkillFed](https://skillfed.io/packages/onnx-weekly) · [View on PyPI](https://pypi.org/project/onnx-weekly/)