Netron is an open-source viewer for neural-network, deep-learning, and machine-learning model files. It provides a visual way to inspect model graphs, operators, tensor shapes, parameters, and metadata without needing to load the model into the original training framework.

The application supports many widely used formats, including ONNX, TensorFlow Lite, PyTorch, TorchScript, torch.export, ExecuTorch, TensorFlow, Core ML, OpenVINO, Keras, Caffe, Darknet, Safetensors, and NumPy, with experimental support for additional ecosystems such as MLIR, JAX, GGUF, PaddlePaddle, and scikit-learn. Netron can be used in a browser, installed as a desktop application, or launched from Python.

Netron is useful to machine-learning engineers, researchers, model-conversion developers, and anyone debugging or validating model artifacts. It is especially convenient when comparing exported models, checking layer connectivity, investigating unexpected tensor dimensions, or understanding an unfamiliar network architecture without writing custom visualization code for each framework.