Deeplearning4j
deeplearning4j.konduit.ai
- Free plan
- Windows
- macOS
- Linux
- API
- Self-hosted
What Deeplearning4j is
Eclipse Deeplearning4j is an open-source deep learning framework for the JVM. It lets developers build, train, and deploy neural networks in Java and Scala, with GPU acceleration through CUDA and CPU computation through OpenBLAS and oneDNN. Apache Spark supports distributed training across CPU or GPU clusters. The framework can import models from Keras, TensorFlow, and ONNX, while Python4J provides Python interoperability from Java through CPython embedding. Its ecosystem includes ND4J, SameDiff, DataVec, Keras Import, Python4J, OmniHub, and Arbiter. Documented deployment settings include JVM microservices, mobile devices, IoT, and Spark environments. The libraries use the Apache License 2.0 and are governed by the Eclipse Foundation. The quickstart targets professional Java developers familiar with production deployments, IDEs, and automated build tools; it recommends Maven, with Ivy and Gradle also listed as options. Documentation covers version 1.0.0-M2.1. Support options include GitHub issues, community forums, Stack Overflow, and professional support from Konduit. The listed open-source plan is free.
Who it suits
Deeplearning4j suits professional Java developers building JVM deep learning applications, including teams that need model import, distributed training, or deployment to microservices, mobile, IoT, or Spark environments.
Good to know, and what to watch
- Builds, trains, and deploys neural networks
- Supports Java and Scala
- Offers CUDA GPU acceleration
- Supports distributed training with Apache Spark
- Imports Keras, TensorFlow, and ONNX models
- Quickstart targets professional Java developers
- Current documented version is 1.0.0-M2.1
Plans
Every plan and priceOpen-source Deeplearning4j
Free plan
Getting started
- Open the Deeplearning4j website
- Create a Java project with Maven, or use Ivy or Gradle
- Add the Deeplearning4j libraries and ecosystem components you need
- Import a Keras, TensorFlow, or ONNX model if applicable
- Build and train the network on CPU, GPU, or an Apache Spark cluster
- Deploy the model in a supported JVM environment
The verdict
Deeplearning4j offers JVM-based model development, training, import, and deployment, including Spark-based distributed training. Its stated audience and Java-oriented quickstart make familiarity with professional Java development relevant.
Questions people ask
Is Deeplearning4j free?
Yes. The open-source Deeplearning4j plan is free and uses the Apache License 2.0.
Which languages does it support?
Supported languages are Java, Scala, Kotlin, and Clojure.
Can it use GPUs and distributed training?
Yes. CUDA provides native GPU acceleration, and Apache Spark supports distributed training on CPU or GPU clusters.
Which model formats can it import?
The suite lists Keras, TensorFlow, and ONNX import support, including Keras H5 and TensorFlow frozen-model files.
Where can models be deployed?
Documented targets include JVM microservices, mobile devices, IoT, and Apache Spark environments.
What support is available?
Support options include GitHub issues, community forums, Stack Overflow, and professional support from Konduit.
What the maker publishes
- Audience
- The quickstart says DL4J targets professional Java developers familiar with production deployments, IDEs, and automated build tools. source
- Commercial support
- Konduit says it provides professional support and software for data science and model serving. source
- Compute
- It provides native GPU acceleration via CUDA and CPU computation via OpenBLAS and oneDNN. source
- Current documentation version
- The homepage says its documentation covers Deeplearning4j 1.0.0-M2.1 as current. source
- Current documented version
- The documentation homepage identifies version 1.0.0-M2.1 as the current version covered. source
- Deployment use cases
- The documentation describes deploying models in JVM microservices, mobile devices, IoT, and Apache Spark environments. source
- Distributed training
- Deeplearning4j supports distributed neural network training on CPU or GPU clusters using Apache Spark. source
- Ecosystem
- Its ecosystem includes ND4J, SameDiff, DataVec, Keras Import, Python4J, OmniHub, and Arbiter. source
- Intended use
- The suite is described for JVM deep learning applications, including importing and retraining models and deploying them in JVM microservices, mobile devices, IoT, and Apache Spark. source
- Languages
- It supports building, training, and deploying neural networks in Java and Scala. source
- License
- The Deeplearning4j project is licensed under Apache License 2.0. source
- Model import
- The documentation lists model import support for Keras, TensorFlow, and ONNX. source
- Model interoperability
- The suite supports importing models from Keras, TensorFlow, and ONNX. source
- Open source
- The libraries are described as completely open source under the Apache 2.0 license and under Eclipse Foundation governance. source
- Purpose
- Eclipse Deeplearning4j is an open-source, distributed deep learning framework for the JVM. source
- Python interoperability
- Python4J provides Python interoperability from Java through CPython embedding. source
- Setup
- The quickstart recommends Maven for Java projects and says other build tools, including Ivy and Gradle, can also work. source
- Spark
- The documentation lists Apache Spark integration for distributed training. source
- Support
- The support page lists GitHub issues, community forums, Stack Overflow, and professional support from Konduit. source
- What it does
- Eclipse Deeplearning4j is an open-source deep learning framework for the JVM, for building, training, and deploying neural networks in Java and Scala. source
Checked against the maker's pages on 8 Oct 2026.