Jupyter Notebook Online: 12 Ways to Run Jupyter in Your Browser, Free and Paid (2026)

You don't need to install anything to use Jupyter. A dozen services run notebooks in the browser, several of them free, and they differ in what matters once you start: how long a session lasts, whether there is a GPU, and whether your files are still there tomorrow.

The short answer: to try Jupyter right now with no account, open JupyterLite at jupyter.org/try — it runs entirely in your browser. For free GPU time, use Google Colab or Kaggle Notebooks; Kaggle publishes its quota, about 30 GPU hours a week. To run a notebook from a public GitHub repository, use Binder. For notebooks a team shares, on the company's data and on a schedule, use a hosted workspace such as Fastero, Deepnote or Hex, or run JupyterHub on your own server.

12 places to run a Jupyter notebook online

Tool	What you get free	Your files	Best for
Google Colab	Free; limits unpublished; sessions up to 12 hours; GPUs when available	Notebooks in Google Drive; the machine itself is wiped	Quick experiments and machine learning
Kaggle Notebooks	12-hour sessions; about 30 GPU hours and 20 TPU hours a week	20 GB of output saved	Free GPUs with published limits
JupyterLite	Runs inside your browser, with no account	Kept in the browser's storage	Trying Jupyter with nothing to install
Binder	1–2 GB of memory, up to 6 hours; stops after 10 minutes idle	Not kept	Running a public repository's notebooks
GitHub Codespaces	120 core-hours a month on personal accounts — about 60 hours on the smallest machine	Kept until deleted, 30 days after stopping	Notebooks inside a code project
Deepnote	3 editors and 5 projects, with Postgres and MySQL connections	Kept	Collaborative notebooks with SQL
Hex	A free Community plan with small machines	Kept	Notebooks that become shareable apps
Databricks Free Edition	Free for non-commercial use, with a daily quota	Kept	Learning Spark and Databricks
Lightning AI Studios	Free credits and one free Studio that restarts every 4 hours	10 GB kept	GPU work in JupyterLab or VS Code
Paperspace Notebooks	A free M4000 GPU, shut down after 6 hours	5 GB kept	A free GPU notebook
CoCalc	2 GB of memory and 3 CPU-hours every 5 hours	Kept	Teaching and live collaboration
Fastero	Notebooks, scheduled runs and free viewers on the Free plan	Kept in your workspace	Teams: notebooks on company data, run on a schedule

The quickest start: JupyterLite and Binder

JupyterLite is JupyterLab compiled to run in the browser: the Python kernel runs on WebAssembly, so there is no server and nothing to sign up for. Open jupyter.org/try and pick JupyterLab or Notebook. The catches: files live in your browser's storage — clearing browser data deletes them — and Python packages that need compiled C code or threads may not work.

Binder turns a public GitHub repository into a temporary Jupyter session: paste the repository's address at mybinder.org and it builds the environment from the repository's requirements.txt. It is a free service run on donated computing, and it has run with reduced capacity since 2023. Each session gets 1–2 GB of memory and up to six hours, stops after 10 minutes of inactivity, and keeps nothing afterwards. Never put secrets in a repository you launch: there is no private launch.

Fastero

Connect your database. Ask questions. Get dashboards.

Postgres, BigQuery, Snowflake, and 10+ sources — live-connected, AI-powered, no dashboard builder learning curve.

Try free →

Free GPUs: Google Colab and Kaggle

Google Colab opens .ipynb notebooks from Google Drive or GitHub in Google's own notebook interface, and is the usual first stop for machine learning. Google is open about one thing: it doesn't publish the limits. Free sessions last at most 12 hours, idle ones time out, and GPUs are "heavily restricted" on the free tier. The machine is recycled, so anything not saved to Drive disappears. In September 2026, Colab became part of Google's consumer AI plans, which add compute units and faster GPUs.

Kaggle Notebooks are the most generous free GPU option with published numbers: about 30 GPU hours and up to 20 TPU hours a week, 12-hour sessions (9 on TPU), and 20 GB of output saved between sessions. Kaggle notebooks can also rerun on a schedule — daily, weekly or monthly — for free. The trade-off is that your work lives on Kaggle, built around its datasets and competitions.

Paperspace Notebooks, now part of DigitalOcean, include a free M4000 GPU on personal accounts, shut down after six hours; free GPUs are in demand during US working hours. Lightning AI Studios give free credits to start and one free Studio, in JupyterLab or VS Code, that has to be restarted every four hours.

Notebooks that are part of your work: Codespaces, Deepnote, Hex and others

GitHub Codespaces opens a repository in a cloud machine with JupyterLab or VS Code. Personal accounts get 120 core-hours a month free; the smallest machine has two cores, so that is about 60 hours. Anaconda Notebooks run JupyterLab with Anaconda's packages and 5 GB of storage on the free plan. CoCalc (now at cocalc.ai) runs .ipynb notebooks in its own collaborative editor, with a shared kernel, edit history and nbgrader for marking assignments.

Deepnote and Hex are data workspaces built around notebooks with SQL and Python, live collaboration and connections to databases. Both save notebooks in their own formats and import .ipynb files; scheduled runs are on their paid Team plans.

Databricks Free Edition, launched in June 2025, replaced the old Community Edition. It is free for non-commercial use with a daily compute quota, and new notebooks there are .ipynb by default.

For a team: notebooks on company data, on a schedule

A notebook that runs in a browser tab is fine for one person exploring. A team needs more: the notebook next to the company's databases, passwords kept out of the code, runs on a timetable, and colleagues who can open the results without installing anything. You can build that with JupyterHub on your own server — The Littlest JupyterHub for one server, Zero to JupyterHub on Kubernetes for more — or use a workspace that does it for you.

In Fastero, notebooks are real .ipynb files — import the ones you have — running Python 3.11 with pandas, SQLAlchemy, seaborn and the Postgres and BigQuery drivers installed. Database passwords and keys come from the workspace's Secrets, not from the notebook. Run a cell, run the whole notebook, or run it on a schedule; colleagues open it with their own login, and viewers are free. The same workspace holds the SQL queries, dashboards and Streamlit apps built from the same data, and Claude can read and run the notebooks through Fastero's connector.

A Jupyter notebook in Fastero: Solar output against forecast, with the code cell run and its table of results, and the project's notebooks, queries and data sources in the explorer

A notebook in Fastero, with the project's queries, apps and data sources beside it. Sample data.

Names you'll see in older lists

Amazon SageMaker Studio Lab no longer accepts new customers; existing users keep it, with no new features.
Databricks Community Edition was retired in 2025 for the Free Edition.
Vertex AI Workbench managed notebooks were deleted on 30 March 2026; Google Cloud now offers Colab Enterprise and Workbench instances instead.
marimo and its cloud service molab are notebooks, with free GPUs in preview, but they use their own Python-file format rather than .ipynb.

How to open an .ipynb file online

To read it: GitHub shows .ipynb files with their saved outputs. Open the file in the repository.
To run it now: in Colab, use File → Upload notebook; in Kaggle, import it into a new notebook; in JupyterLite, drag it into the file browser.
To run it from a public repository: paste the repository into mybinder.org.
To run it with a team: import it into a workspace such as Fastero, Deepnote or Hex, where colleagues can open it too.

FAQ

Can I run Jupyter Notebook online for free?

Yes. JupyterLite at jupyter.org/try runs Jupyter in your browser with no account. Google Colab and Kaggle Notebooks are free with a Google or Kaggle account and include GPU time, and Binder runs notebooks from public GitHub repositories for free.

Which online Jupyter service has a free GPU?

Kaggle Notebooks give about 30 GPU hours a week and publish the quota. Google Colab's free tier includes GPUs when they are available, with limits Google doesn't publish. Paperspace offers a free M4000 GPU on personal accounts, and Lightning AI gives free credits to spend on GPUs.

Is Google Colab the same as Jupyter Notebook?

Colab is built on Jupyter and stores notebooks in the standard .ipynb format, but it uses Google's own interface instead of JupyterLab, stores notebooks in Google Drive, and runs them on Google's machines with its own limits.

Do online Jupyter notebooks save my files?

It depends. Binder and JupyterLite keep nothing on a server; Colab keeps notebooks in Google Drive but wipes the machine; Kaggle saves 20 GB of output; Codespaces, Deepnote, Hex, CoCalc and Fastero keep your files in your account or workspace.

How can a team share Jupyter notebooks online?

Use a shared workspace — Fastero, Deepnote, Hex or Databricks — or run JupyterHub on your own server. Look for sign-in for colleagues, a place for database passwords outside the notebook, and scheduled runs if the notebook produces a regular report.

Try Fastero free — run your Jupyter notebooks next to your team's data, on a schedule, shared with a sign-in. No credit card required.