Google Colab is a low barrier option to run Python scripts. Here is a brief introduction for the essential features.

For who is Google Colab suitable for?

Golab is simple solution to run arbitrary Python scripts and collaborate with large group of people for free. Education is a good example.

Sharing the notebooks is easy, just like you would share files in Google Drive. Google Colab is for Python the same as Google Sheets is for spreadsheet.

Notebooks can be loaded from GitHub. This feature can be found from File > Upload > GitHub.

Colab provides also GPU free of charge, which is quite cool to train your machine learning skills. This has been considered as Colab’s biggest edge in discussion forums.

Colab free subscription comes with privacy questions

Google possibly take advantage of leftover compute capacity to make the solution cost efficient. They say that the available resources vary over time.

How is any other Google product free? At least I have some privacy concerns.

Even though, code is executed on a user-specific virtual machine and deleted after a session. This can be seen as a positive act to prevent user data leaks.

Read here the most frequent Colab questions .

Colab Pro ja Pro+

Pro version comes with 100 compute units for 11,5 euros per month. There is no clear definition for a “unit”.

While Colab free instances can run for 12 hours max, Pro+ version extends this period to 24 hours. Pro+ is 52 euros per month for 500 compute units.

Both enable terminal for low level commands.

Paid version is available only in specified countries.

Here is the Colab pricing page .

Google Colab issues with larger apps

Here is my honest opinion.

Colab is not suitable for serious data science practitioners. Computation can not be scaled and usage limits fluctuate.

Google Drive is your only storage option. This possibly limits the data size. It is also a governance issue for larger companies. The documentation says that Drive mounts might fail with too many sub folders.

Larger applications would be near impossible to build with Colab. There is no easy solution to even import other Python files to organize the code.

Google Colab alternatives

For simple scripts, maybe CoCalc, Python Anywhere and PyScripts would be the closest alternatives.

For professional data science teams there better options. Anaconda Cloud, Datalore and Vertex AI to mention a few.

At the end of June it was announced that Google Colab Enterprise will be integrated to Vertex AI in Google Cloud. It remains to be seen what this means in long term.