Jupyter Notebook platforms
We have compared 26 Jupyter Notebook (JN) platforms against 11 criteria, to decide which platforms to support. Although we selected criteria that are relevant to our use case, writing and sharing algorithmic essays, we hope they are general enough to be of use by anyone considering the adoption of JN platforms. A comparison table is provided at the bottom of this page.
The datasciencenotebook.org website by Robert Lacok and the awesome-jupyter GitHub repository started by Markus Schanta were invaluable in locating and evaluating many of these platforms.
Evaluation criteria
We used the following criteria to evaluate the platforms:
- Full compatibility with JN files (.ipynb)
- Free account
- Hosting mode (locally or in the cloud)
- Customisable run-time environment
- Share notebooks via links with access levels
- GitHub integration
- User interface (classic or bespoke)
- Real time collaboration
- Commenting on cells
- Access to terminal and file system
- Access to underlying notebook files
The following paragraphs briefly note why we have chosen each aspect, and why these features are desirable in educational settings and for writing algorithmic essays.
An important consideration for any platform is whether it is fully compatible
with .ipynb files. This file type, used by many platforms, originates from
when the platform was called IPython Notebook. The ability to import and export
these files removes dependency on a single JN platform, allowing to write an essay
in one platform and share it on another.
Being free to use, or having a free account, is a highly desirable feature of any platform. It lowers the barrier to participation for both students and educators, removes extra budget requirements, and can increase the longevity of projects.
The hosting mode (locally or in the cloud) has implications for setup requirements, UI, and collaboration. Cloud-based platforms require the least setup and typically support collaboration and commenting on cells. However, they mostly cater for data scientists in their documentation, and the UI and feature set can change without notice. Access to the underlying notebook files may be restricted, complicating code linting, and interfaces can vary, increasing the learning curve. In contrast, locally installed JN platforms tend to have a familiar and consistent interface, provide unrestricted access to files and have more balanced documentation. However, many lack simple options for collaboration and commenting, and they must be installed and configured by the user.
Any platform suitable for pedagogical practice and particularly writing algorithmic essays must have a customisable run-time environment. It should be adaptable to different course specifications, allowing the use of third-party libraries and tools. Often, the ability to install new Python packages via pip is sufficient, but some platforms go further by allowing environment customisation through a Dockerfile.
Students are often required to share their work with faculty and peers for marking, receiving feedback, and collaboration. If a platform has the option to share notebooks via links with access levels, students can grant recipients permission to read only, leave comments on, or edit the document based on their role. Links can be sent by email to specific individuals or shared with a wider audience on a forum. In some cases, work can be shared without the recipient needing an account on the same platform.
The ability for a JN platform to integrate with other platforms can be highly desirable. For instance, integration with GitHub allows one to leverage version control when developing resources, use a single repository to serve multiple platforms, and back up materials.
The user interface is often the first element of a platform a user encounters. Ease of use, familiarity and which controls are available are all important factors that can help reduce the learning curve. For users already familiar with Jupyter, it’s important to consider whether a platform provides the classic JN interface or a bespoke interface. For the purpose of this comparison, we do not consider Jupyter Lab to have the classic interface.
Some platforms allow real-time collaboration among their users, to work together on documents simultaneously, and see changes instantly. Collaborating this way fosters teamwork and communication skills.
Commenting on cells is useful in many ways. Comments are located close to the relevant code or text, often contained in separate “threads” where they can be edited, deleted and resolved. This allows other users to provide feedback (and educators to mark) without modifying the existing structure of the notebook.
Having access to a terminal and the underlying file system allows platform users to run custom scripts and manage auxiliary files.
Having access to the underlying notebook files in combination with access to a terminal, allows linters and formatters to be used on these files.
Comparison table
You may need to scroll sideways to see all columns. You can sort the table by clicking on any column header.
Some platforms may no longer exist, or may have changed their name or URL, since we compared them in January 2024.
| Platform | .ipynb Compatibility | Free Account | Hosting | Customisable Run-time | Share via Links | GitHub Integration | Interface | Real-time Collaboration | Commenting on Cells | Terminal & File System Access | Access to Notebook Files |
|---|---|---|---|---|---|---|---|---|---|---|---|
| Amazon SageMaker Notebooks | ✓ | ✓ | Cloud | ✓ | ✓ | ✓ | Bespoke | ✓ | ✓ | ✓ | ✓ |
| Apache Zeppelin | ✓ | ✓ | Local/Cloud | ✓ | ✓ | ✓ | Bespoke | ✓ | ✓ | ✓ | ✓ |
| Azure Notebooks | ✓ | ✓ | Cloud | ✓ | ✓ | ✓ | Bespoke | ✓ | ✓ | ✗ | ✗ |
| Carnets (iOS) | ✓ | ✓ | Local | ✓ | ✗ | ✗ | Classic | ✗ | ✗ | ✓ | ✓ |
| CoCalc | ✓ | ✓ | Cloud | ✓ | ✓ | ✓ | Bespoke | ✓ | ✓ | ✓ | ✓ |
| CodeOcean | ✓ | ✓ | Cloud | ✓ | ✓ | ✓ | Bespoke | ✓ | ✓ | ✓ | ✓ |
| Count | ✗ | ✓ | Cloud | ✓ | ✓ | ✓ | Bespoke | ✓ | ✓ | ✗ | ✗ |
| Datalore (JetBrains) | ✓ | ✓ | Cloud | ✓ | ✓ | ✓ | Bespoke | ✓ | ✓ | ✓ | ✓ |
| Deepnote | ✓ | ✓ | Cloud | ✓ | ✓ | ✓ | Bespoke | ✓ | ✓ | ✓ | ✗ |
| Google Colab | ✓ | ✓ | Cloud | ✓ | ✓ | ✓ | Bespoke | ✓ | ✓ | ✗ | ✗ |
| Hyperquery | ✗ | ✓ | Cloud | ✓ | ✓ | ✓ | Bespoke | ✓ | ✓ | ✗ | ✗ |
| IBM Watson Studio | ✓ | ✓ | Cloud | ✓ | ✓ | ✓ | Bespoke | ✓ | ✓ | ✗ | ✗ |
| Jupyter Notebook | ✓ | ✓ | Local | ✓ | ✗ | ✓ | Classic | ✗ | ✗ | ✓ | ✓ |
| JupyterLab | ✓ | ✓ | Local | ✓ | ✗ | ✓ | Bespoke | ✓ | ✗ | ✓ | ✓ |
| Kaggle Notebooks | ✓ | ✓ | Cloud | ✗ | ✓ | ✓ | Bespoke | ✓ | ✓ | ✗ | ✗ |
| Mode Notebooks | ✗ | ✓ | Cloud | ✓ | ✓ | ✓ | Bespoke | ✓ | ✓ | ✗ | ✗ |
| Noteable | ✓ | ✓ | Cloud | ✓ | ✓ | ✓ | Bespoke | ✓ | ✓ | ✗ | ✗ |
| nteract | ✓ | ✓ | Local | ✓ | ✗ | ✓ | Classic | ✗ | ✗ | ✓ | ✓ |
| Observable | ✗ | ✓ | Cloud | ✗ | ✓ | ✗ | Bespoke | ✓ | ✓ | ✗ | ✗ |
| Polynote | ✗ | ✓ | Local | ✓ | ✓ | ✓ | Bespoke | ✓ | ✓ | ✓ | ✓ |
| Pluto.jl | ✗ | ✓ | Local | ✓ | ✗ | ✗ | Bespoke | ✗ | ✗ | ✓ | ✓ |
| Query.me | ✗ | ✗ | Cloud | ✓ | ✓ | ✓ | Bespoke | ✓ | ✓ | ✗ | ✗ |
| Quarto | ✓ | ✓ | Local | ✓ | ✗ | ✓ | Classic | ✗ | ✗ | ✓ | ✓ |
| Saturn Cloud | ✓ | ✓ | Cloud | ✓ | ✓ | ✓ | Bespoke | ✓ | ✓ | ✓ | ✓ |
| VS Code with Jupyter | ✓ | ✓ | Local/Cloud | ✓ | ✗ | ✓ | Classic/Bespoke | ✓ | ✓ | ✓ | ✓ |
| Zepl | ✗ | ✓ | Cloud | ✓ | ✓ | ✓ | Bespoke | ✓ | ✓ | ✗ | ✗ |