Direct comparison
Google Colab vs Jupyter Notebook: Where Code Runs
Google Colab vs Jupyter Notebook for research: who runs the server, runtime limits, what a shared notebook carries, and what data leaves your machine.
Written and maintained by CASRAI Editorial Board
Last updated
Ask CASRAI · free to try
Ask about Google Colab vs Jupyter Notebook: Where Code Runs
Ask your first 2 questions free below. Subscribers get 150 a day for $29 a month.
An AI assistant specialized in research administration. It cites the sources behind every answer, labels web answers and says when it can't answer.
Answers draw on CASRAI's guides and dictionary plus the federal and funder documents we index: Federal Register, Grants.gov, Regulations.gov and UKRI.
Works on this site and inside Claude, Cursor and the AI tools you already use.
Everything CASRAI publishes — this page, the dictionary, the guides and the news — stays free to read, with no account and no card.
How do Google Colab, Jupyter Notebook compare side by side?
The table below compares Google Colab, Jupyter Notebook across 10 procurement-relevant dimensions, from what it is through source.
Side-by-side comparison
| Dimension | Google Colab | Jupyter Notebook |
|---|---|---|
| What it is | A hosted Jupyter Notebook service that requires no setup to use, built on open-source Jupyter. | A notebook authoring application under the Project Jupyter umbrella. |
| Who runs the server | Google. Notebooks run on virtual machines that the Colab service creates and deletes. | You. A URL starting http://localhost: or http://127.0.0.1: means your computer is acting as the server; your company or university might instead run the server for you. |
| Setup and cost | No setup. Colab is free of charge to use; Colab Pro, Pro+ and Pay As You Go offer increased compute availability based on your compute unit balance. | You install and start the application yourself (default port 8888). The software is released under the BSD 3-Clause licence. |
| Runtime limits | Resources are not guaranteed and not unlimited, and usage limits sometimes fluctuate. Free notebooks can run for at most 12 hours; Pro+ allows continuous execution for up to 24 hours given enough compute units. Idle runtimes time out and virtual machines have a maximum lifetime. | The kernel keeps running after you close the browser window. Limits come from the machine or server you use, not from the application. |
| Sharing | Shared as you would share Google Docs or Sheets: text, code, output and comments are shared. The virtual machine, its custom files and libraries are not shared, so include install cells. | Notebooks are .ipynb files, which are JSON internally, so they can be version-controlled and shared as files. |
| Reproducibility | A new runtime starts without the previous machine's files and libraries, so the notebook itself must reinstall what it needs. | One notebook uses one kernel session on your machine; the environment is whatever you have installed there. |
| Data handling | For generative AI features Google collects prompts, related code, generated output, usage information and feedback, and human reviewers may read, annotate and process them. A notebook setting, Omit code cell output when saving this notebook, is available under Edit > Notebook settings. | The documentation states that Jupyter does not send your data anywhere else. Unsigned notebooks do not display JavaScript or HTML output until you trust them. |
| Using your own machine | A local runtime can be attached, using your own compute. Google Drive mounting does not work with that approach. | This is the default way to run it, locally or on a server you control. |
| Pick it when | You want to open a notebook in a browser with nothing installed and can live with limits that are not guaranteed. | You want the kernel on infrastructure you control, with no hosting-service session cap and no data sent to a third party. |
| Source | Colab FAQ, https://research.google.com/colaboratory/faq.html (read 2026-10-29). | Jupyter Notebook documentation, https://jupyter-notebook.readthedocs.io/en/stable/notebook.html (read 2026-10-29); licence, https://raw.githubusercontent.com/jupyter/notebook/main/LICENSE (read 2026-10-29). |
Common questions
Common questions about Google Colab vs Jupyter Notebook
Is Google Colab the same as Jupyter Notebook?
+
Not exactly. The Colab FAQ describes Colab as a hosted Jupyter Notebook service that requires no setup and is built on open-source Jupyter. Jupyter Notebook is the application you run yourself. They share the notebook idea; the difference is who runs the server.
Colab vs Jupyter: which is better for research?
+
It depends on where the work needs to run. Colab removes setup but uses Google-run virtual machines with limits that are not guaranteed. Jupyter Notebook keeps the kernel on a machine you or your institution control. See the rows on runtime limits and data handling above.
Jupyter vs Colab: how long can a notebook run?
+
Colab says free notebooks can run for at most 12 hours, and Pro+ allows continuous execution for up to 24 hours with sufficient compute units; runtimes also time out when idle. With Jupyter Notebook the kernel stays active even if the browser window is closed, and any limit comes from the machine you use.
Can Google Colab use my own computer?
+
Yes. The Colab FAQ says you can use your own compute through a local runtime you control. Google Drive mounting does not work with these approaches.
Does Jupyter Notebook send my data anywhere?
+
The Jupyter Notebook documentation states that Jupyter does not send your data anywhere else. If you run the server on your computer, the URL starts with http://localhost: or http://127.0.0.1:.
