Magika

A state-of-the-art AI tool designed to quickly identify file types. Magika uses a lightweight deep learning model to analyze file contents and accurately identify over 200 different content types, including binary formats, source code, documents, and data science files, even when extensions are missing or misleading. It achieves around 99% average precision and recall on diverse test sets, often outperforming older signature-based tools by 20% or more, especially on tricky textual formats like code or config files. The model is highly optimized at just a few MB and runs efficiently on a single CPU, with inference typically around 5ms per file after the initial load. It supports 200+ types, covering everything from common formats like PDF, JPEG, Python scripts, and Excel to specialized ones like Jupyter notebooks, PyTorch models, Dockerfiles, TOML, and various programming languages that traditional detectors often confuse. The code, model, and bindings are open source under Apache 2.0, available on GitHub, with easy installation via pip for Python or as a Rust CLI. Google provides a web demo that runs entirely in your browser, letting you upload files and see real-time identifications powered by the JavaScript/TypeScript binding.