---
id: tf-keras-nightly
version: "2.21.0.dev2026081409"
license: Apache 2.0
license_treatment: permissive
maintenance: active
---

# tf-keras-nightly — Deep learning for humans.

License: permissive · Maintenance: active · Downloads: 1.4M/mo

## What it is and what it does

TF-Keras is the pure-TensorFlow implementation of Keras, providing a high-level API for defining, training, and evaluating deep learning models. It is built on the legacy tf.keras codebase and runs entirely within TensorFlow, making it tightly integrated with TensorFlow's execution model and optimization pipeline. The package offers familiar Keras abstractions—layers, models, optimizers, and loss functions—while remaining a TensorFlow-native implementation.

This is a nightly development build, meaning it tracks the latest development branch of TF-Keras and receives frequent updates. It depends on tf-nightly, so you get the latest TensorFlow features and fixes alongside the latest Keras API changes. This package is distinct from the multi-backend Keras 3, which supports multiple backends; TF-Keras is TensorFlow-specific and maintains the original Keras API surface for TensorFlow users.

Use it for:

- Build neural network models for image classification, NLP, or regression tasks using the Keras API.
- Experiment with cutting-edge TensorFlow and Keras features during development by tracking nightly builds.
- Train models with custom layers, callbacks, and training loops while leveraging TensorFlow's distributed training.
- Migrate existing tf.keras code to a standalone package that decouples Keras from the main TensorFlow release cycle.
- Prototype deep learning architectures using high-level Keras abstractions without writing low-level operations.

## Worth the install?

AI-flagged interpretation of the facts above — verify before relying.

TF-Keras is a pure-TensorFlow implementation of Keras that provides a high-level deep learning API for building and training neural networks.

Yes, if you are actively developing with TensorFlow and want the latest Keras API features; no, if you need stability and predictability—use the stable tf-keras release instead. The nightly build is suitable for developers and researchers who can tolerate frequent changes and want early access to new features, but not for production systems requiring long-term API stability.

## Install

pip install tf-keras-nightly
uv add tf-keras-nightly
poetry add tf-keras-nightly

## Installing tf-keras-nightly

Before you install: Low install friction with a pure Python wheel. Active maintenance with a commit from 2026-08-14 and a release on the same day. Single runtime dependency on tf-nightly means you inherit that package's stability profile.

License in practice: Apache 2.0 is permissive; you can use this in commercial and proprietary projects with minimal restrictions, provided you retain license notices.

Quickstart:

pip install tf-keras-nightly

from tf_keras_nightly import keras

model = keras.Sequential([
    keras.layers.Dense(activation='relu'),
    keras.layers.Dense(activation='softmax')
])
model.compile(optimizer='adam', loss='sparse_categorical_crossentropy')

Requires Python 3.10 or later and tf-nightly as a runtime dependency.

Verify before relying:

- Whether this nightly build is suitable for production use or intended only for development and testing
- Stability guarantees and API compatibility between nightly releases
- Performance characteristics compared to the stable tf-keras release channel
- Specific usage examples and API surface details beyond what the description excerpt provides

## Package facts

- License: Apache 2.0 (permissive)
- Python support: supports_current
- Install friction: low
- Maintenance: active
- Downloads: 1.4M/month (top 5,000 on PyPI)
- Known vulnerabilities: none known

## Tags

keras tensorflow deep learning, neural network api tensorflow, deep learning framework python, tensorflow keras layers models, machine learning model building, tensorflow high-level api, deep-learning, nightly-build, tensorflow-native

[View on SkillFed](https://skillfed.io/packages/tf-keras-nightly) · [View on PyPI](https://pypi.org/project/tf-keras-nightly/)