tfjs and tfjs-core
TensorFlow.js is a high-level library that depends on tfjs-core as its underlying engine, providing the WebGL acceleration and autodiff primitives while tfjs-core handles the low-level tensor operations and GPU computation.
About tfjs
tensorflow/tfjs
A WebGL accelerated JavaScript library for training and deploying ML models.
Supports both browser and Node.js execution through pluggable backends (WebGL, WebAssembly, WebGPU, and native TensorFlow C++), enabling client-side inference and retraining with sensor data. Provides layered APIs from low-level linear algebra operations to high-level Keras-compatible layers, plus tools to import pre-trained TensorFlow and Keras models. Includes specialized packages for data processing, visualization, and AutoML model loading, with individual module imports available to minimize bundle size.
About tfjs-core
tensorflow/tfjs-core
WebGL-accelerated ML // linear algebra // automatic differentiation for JavaScript.
Provides GPU-accelerated tensor operations and computational graph differentiation through WebGL, with CPU fallback support. Implements eager execution for dynamic computation graphs while maintaining automatic gradient computation for training neural networks. Integrates as the core numeric engine for TensorFlow.js, enabling in-browser ML inference and training across multiple backends.
Related comparisons
Scores updated daily from GitHub, PyPI, and npm data. How scores work