Loading...
Backed By Keras Team
High-level neural networks API for fast deep learning prototyping.
Keras is an open-source, high-level neural networks API written in Python. It runs on top of TensorFlow and other frameworks, enabling fast prototyping, easy model building, and deployment. Keras is widely used for deep learning applications in research and industry, supporting convolutional and recurrent networks, transfer learning, and model export for production.
Fast model building
Supports CNNs, RNNs, and more
Export models to TensorFlow
Deploy to cloud or edge
Leverage pre-trained models
Fine-tune for custom tasks
Full access to all Keras APIs and functionalities on all supported backends (JAX, TensorFlow, PyTorch).
**Multi-Backend Support (Keras 3.x):** Supports JAX, TensorFlow, and PyTorch (and OpenVINO for inference).
Sequential API, Functional API, and Model Subclassing for all backends.
The **`keras.ops`** namespace for writing backend-agnostic custom components.
Backend-Agnostic Model Saving/Loading (`.keras` format).
Compatibility with native ecosystem tools (e.g., Keras models work as PyTorch Modules or JAX functions).
Real user experiences from across different platforms
Keras 3 is a multi-backend deep learning framework, with support for JAX, TensorFlow, PyTorch... Effortlessly build and train models for computer vision, natural language processing, audio processing, timeseries forecasting... State-of-the-art performance: By picking the backend that is the fastest for your model architecture (often JAX!), leverage speedups ranging from 20% to 350%...
Keras Team (GitHub)
2025 (latest releases)
-art performance and scalability without being locked into a single framework.
on-ready high-level API.
ons (requires Python 3.10+).
Unprecedented framework flexibility, allowing users to switch backends to achieve optimal performance (e.g., JAX for speed) or deployment compatibility (e.g., TensorFlow for TFLite) without rewriting code. It maximizes developer productivity and code readability.
Requires **Python >=3.10** for Keras 3.x (Source 2.2). Switching the backend requires configuring the environment variable *before* importing Keras. Custom components relying on framework-specific code may break cross-backend compatibility.