JupyterLab: The essential tool for AI developmentJupyterLab: The essential tool for AI development

JupyterLab is the latest and improved version Jupyter Notebook, an essential tool for ML engineers and data scientists. It is more than just an IDE for AI developers — it is a flexible, web-based interface that combines code execution, rich text, visualizations and more into a unified workspace that adapts to your workflow.

JupyterLab provides the perfect environment for quick data analysis, ad-hoc prototyping, fine-tuning and model evaluation. Today, JupyterLab stands as the de facto standard for interactive computing and experimentation in the AI world. It allows you to develop, test and refine AI models with immediate feedback, process documents along with your code, collaborate and share discoveries seamlessly with colleagues.

Running JupyterLab on Nebius AI CloudRunning JupyterLab on Nebius AI Cloud

We received many requests from our users asking us to make JupyterLab available as a managed cloud application — without the need to deploy and configure a Kubernetes cluster underneath. Therefore, JupyterLab runs in our Standalone Applications service, providing a smooth and simplified user experience where a user can start working in a notebook without even touching the infrastructure management layer.

An additional significant benefit of this service is the capability to run JupyterLab on NVIDIA GPUs. This means that you can start developing and prototyping your ML tasks and launch a lightweight GenAI training right away. No need to provision compute or install drivers — just select the proper configuration (1x NVIDIA H100 GPU, 8x NVIDIA H100 GPUs or CPU-only VMs available) and launch the application.

Features and limitations of the release versionFeatures and limitations of the release version

While this first version has some limitations in terms of time for the first run and number of instances of the JupyterLab running simultaneously, we’ve designed it to meet the needs of most development workflows. As a user you will get:

• Pre-configured PyTorch® framework.
• Persistent storage, to save your progress between sessions.
• Easy package management, to customize your environment.
• Ability to pause the application, to control compute costs.
• The hourly price for your JupyterLab app starts from $2.95 per hour for the NVIDIA H100 GPU configuration. Persistent storage is charged additionally.

What’s nextWhat’s next

This release is just a beginning of Nebius building a comprehensive AI development toolset that supports the entire machine learning lifecycle. Here are some of the ideas we plan to deliver later this year:

• Additional AI applications with the same streamlined experience.
• Enhanced integration between applications, for smoother workflows.
• More customization options to tailor environments to your specific needs.
• Expanded collaboration features to better support team-based development.

We believe that by removing friction from the development process, we can help AI professionals focus on what matters most: solving important problems and building innovative solutions.

Getting startedGetting started

You can access JupyterLab in the Standalone Applications service in our cloud console — just sign up for the platform and navigate to the Applications section. We would also appreciate your feedback about this app, so please let us know what you think. Your insights and experience can help us refine and improve the product offering and shape the future of our platform.