#8383rd of 105 in ML frameworks and libraries
flax
Neural network library for JAX built around mutable Python objects
- Stars
- 7.3k
- License
- Apache-2.0
- Last commit
- Oct 2026
- Last release
- Sep 2026
- Language
- Jupyter Notebook
Missed the latest GitHub check; these facts may be older.
Overview
Flax is a neural network library for JAX. Its current API, Flax NNX, lets you define models as regular Python objects with reference sharing and mutability, and ships layers such as Linear, Conv, BatchNorm, LayerNorm, MultiHeadAttention, LSTMCell and Dropout. It also includes utilities for replicated training, checkpointing and metrics, plus MNIST and Gemma inference examples.
Who it is for: Researchers and engineers building and training neural networks in JAX
Strengths
- Flax NNX models are plain Python objects, so they can be inspected and debugged directly
- Developed with the JAX team; runs on CPU, GPU and TPU through JAX
- Includes common layers plus checkpointing, serialization, metrics and device prefetching
- Built for forking examples and editing the training loop, not extending a framework
Weaknesses
- Two APIs exist, Linen and NNX, with separate docs sites and a migration guide
- Requires JAX and its own accelerator setup, which Flax does not handle
- Only MNIST and Gemma inference are listed as built-in examples
- Described as not an official Google product
What it needs
- GPU optional
- Needs JAX, Python 3.8+
- Models: Gemma (inference example), MLP, CNN, autoencoder
Also in ML frameworks and libraries
See all 105| Rank | Project | Score |
|---|---|---|
| 1 | transformersPython library defining open models for text, vision, audio and multimodal tasks | 94 out of 100 |
| 2 | pytorchPython tensor library with GPU support and tape-based autograd for deep learning | 94 out of 100 |
| 3 | ultralyticsPython library and CLI for training and running YOLO vision models | 93 out of 100 |
| #4 | pytorch-image-modelsPyTorch library of image models, pretrained weights and reference training scripts | 92 out of 100 |
| #5 | supervisionPython toolkit for loading, annotating and counting computer vision detections | 91 out of 100 |