Is Kaggle Notebooks free?

1. Create a notebook for your first test

Open Kaggle’s Code page, sign in and select New Notebook. Give it a name you will recognize, such as ‘My first Kaggle notebook’. A notebook combines code cells, their results and written notes in your browser. Kaggle runs the code on its remote machine; installing Python on your laptop is unnecessary for this exercise.

Use a new notebook rather than editing somebody else’s project. Keep it private while learning, and check sharing before adding files. This example needs no uploads, datasets, API keys or paid plan. The walkthrough uses a desktop layout; menus may move on a narrower screen.

1

2. Choose CPU first; find GPU when you need it

In the editor, open Settings, then Accelerator. Leave None selected for the starter code below: None means no GPU or TPU accelerator, so ordinary Python still runs on CPU. Our menu offered GPU T4 x2 and TPU v5e-8. Those are observed choices, not a promise of the same hardware or immediate availability for every account.

Choose a GPU later for code that actually uses it, such as an appropriately configured PyTorch or TensorFlow workload. Selecting a GPU does not automatically make every Python program faster. Avoid using GPU time for editing text, preparing files or this small arithmetic example.

Kaggle’s efficient-usage documentation describes a weekly GPU quota of 30 hours, sometimes higher depending on demand and resources. Check your own current quota and eligibility before planning a long run. The documentation and interface can show different hardware generations; use the actual selectable option for your run.

123

3. Run a cell and check its result

Replace the starter cell with the Python below. Click its triangular play button or Run current cell in the toolbar. Wait for the session to start and the cell to finish. You should see ‘FreeAIList example: average = 5.0’ and ‘Saved: /kaggle/working/freeailist-example.txt’ below the cell. We observed both lines in our CPU test.

The code averages four numbers and writes the result to a small text file. It uses Python’s standard library and makes no network requests. This checks the basic notebook workflow; it is not an AI training example or a GPU performance benchmark. For an AI project, first confirm its package, memory, dataset and model-license requirements separately.

from pathlib import Path

values = [2, 4, 6, 8]
average = sum(values) / len(values)
result = f"FreeAIList example: average = {average:.1f}"
print(result)

Path("/kaggle/working/freeailist-example.txt").write_text(result + "\n")
print("Saved: /kaggle/working/freeailist-example.txt")

12

4. Save your work and choose whether to keep files

Select Save Version at the top right. The dialog defaults to Save & Run All (Commit), which starts a fresh top-to-bottom execution. For this already-run exercise, switch Version Type to Quick Save. It snapshots the notebook as it currently looks, so you do not need to rerun the code simply to save a version.

Open Advanced Settings and inspect Save Output. In our dialog it defaulted to ‘Never save output when creating a Quick Save’. To retain the generated text file, choose ‘Save output for this version when creating a Quick Save’ before selecting Save. Wait for completion. Our saved Version 1 completed with output enabled. The screenshot shows a later inspection of the default setting; that second dialog was canceled.

Writing a file to /kaggle/working during a session and preserving it in a saved version are different actions. Save or download files you need before ending the session. Use Save & Run All deliberately when you want to check that every cell works from a clean state; do not launch repeated background runs as a substitute for saving edits.

123

5. Stop the session before closing the tab

Once the version and any files you need are safely saved, open the three-dot More settings menu beside Draft Session and select Stop session. Check that Draft Session changes to ‘off (run a cell to start)’. We stopped this walkthrough’s session and confirmed that status. Interrupt execution only stops running code; use Stop session to shut down the session itself.

Closing the browser tab or restarting your laptop does not reliably stop a remote notebook. Kaggle’s efficient-usage guide specifically recommends stopping interactive sessions before closing the window. Check Active Events for work you intentionally left running, and stop only the session you mean to end.

12

• The version has finished saving, and required files are preserved or downloaded.
• The correct notebook’s Draft Session says off.
• No unnecessary Save & Run All jobs were started for this exercise.

If something looks different or does not work

If GPU choices are unavailable, read the message shown in your account. Possible issues include verification, quota or resource availability. This test used an existing account; it does not establish that a new account can immediately start a GPU. Follow Kaggle’s own verification flow rather than entering payment details into an unrelated offer.

If the cell is queued or the session is starting, wait for its status to change before clicking Run again. If Python shows an error, compare the pasted code, quotation marks and file path with the example. A successful screenshot from our account is a reference, not a guarantee of a particular startup time for yours.

For longer projects, save checkpoints and important files deliberately. A draft, an interactive session and a saved version serve different purposes. Read the current runtime limits and your account’s remaining quota before leaving a training job unattended. We did not test long runs or train a model in this walkthrough.

When to choose Google Colab instead

Consider Colab if you prefer its notebook interface or already organize your work around Google Drive. It is another browser route for running Python and eligible accelerator workloads.

Colab’s free resources, hardware and usage limits vary. Switching does not guarantee a GPU or a fixed uninterrupted runtime. Compare the current restrictions and save your work whichever service you choose.