# Netron

> Netron is an open-source viewer for neural network, deep learning, and machine learning models, supporting ONNX, TensorFlow, PyTorch, and many more formats.

Netron is an open-source model visualizer created by Lutz Roeder, available as a browser app, desktop application, and Python package. It renders the computation graphs of neural network and machine learning models, making it easy to inspect layer structure, tensor shapes, and operator attributes. The project is hosted on GitHub under the MIT License and has accumulated over 32,000 stars as of mid-2026.

## What It Is

Netron is a graph-based viewer that loads a saved model file and renders its architecture as an interactive node-and-edge diagram. It targets ML practitioners, researchers, and engineers who need to understand, debug, or document model structure without writing custom inspection code. The tool operates entirely client-side in the browser or as a native desktop app, meaning no model data is sent to a remote server.

## Format Support

Netron covers a wide range of model serialization formats across the major ML frameworks:

- **Stable support**: ONNX, TensorFlow Lite, PyTorch, torch.export, ExecuTorch, TorchScript, TensorFlow, Core ML, OpenVINO, Keras, Caffe, Darknet, Safetensors, NumPy
- **Experimental support**: MLIR, JAX, GGUF, RKNN, ncnn, MNN, PaddlePaddle, scikit-learn

This breadth makes Netron useful across research, production, and edge-deployment workflows regardless of the originating framework.

## Deployment Options

Netron can be accessed in several ways:

- **Browser**: The hosted web app at netron.app runs entirely in the browser with no installation required.
- **macOS**: Distributed as a `.dmg` file or via `brew install --cask netron`.
- **Linux**: Available as `.deb` or `.rpm` packages.
- **Windows**: Distributed as an `.exe` installer or via `winget install -s winget netron`.
- **Python**: Installable via `pip install netron`, then invoked with `netron [FILE]` or `netron.start('[FILE]')` from within a script or notebook.

## Update: v9.0.9

The latest release is version 9.0.9, published on May 23, 2026. The repository was last pushed to on May 25, 2026, indicating active ongoing maintenance. The project has been continuously developed since its creation in December 2010, with the primary language being JavaScript. The GitHub repository lists 19 open issues and over 3,100 forks, reflecting broad community engagement.

## Why It Matters for ML Workflows

Model visualization is a practical necessity when debugging unexpected outputs, verifying export correctness after framework conversion (e.g., PyTorch → ONNX), or communicating architecture decisions to collaborators. Netron fills this role without requiring a running training environment — a user only needs the saved model file. Its support for Safetensors and GGUF also makes it relevant for the growing ecosystem of large language model weights distributed via Hugging Face and similar platforms.

## Features
- Interactive neural network graph visualization
- Support for ONNX, TensorFlow, PyTorch, Core ML, Keras, and more
- Experimental support for GGUF, JAX, MLIR, and scikit-learn
- Browser-based viewer with no server-side model upload
- Desktop apps for macOS, Linux, and Windows
- Python package with programmatic API
- Inspect layer attributes, tensor shapes, and operator details
- Open model files via URL in the browser version

## Integrations
ONNX, TensorFlow, TensorFlow Lite, PyTorch, TorchScript, ExecuTorch, torch.export, Core ML, OpenVINO, Keras, Caffe, Darknet, Safetensors, NumPy, MLIR, JAX, GGUF, RKNN, ncnn, MNN, PaddlePaddle, scikit-learn, Hugging Face

## Platforms
WINDOWS, MACOS, LINUX, WEB, API, CLI

## Pricing
Open Source

## Version
v9.0.9

## Links
- Website: https://netron.app
- Documentation: https://github.com/lutzroeder/netron
- Repository: https://github.com/lutzroeder/netron
- EveryDev.ai: https://www.everydev.ai/tools/netron