Notebooks and ML

16 services

Hosted notebooks are limited by runtime hours, hardware and session length rather than storage. Kaggle gives 30 GPU hours and 20 TPU hours a week, and its CPU sessions get 4 cores and 30 GB of memory. Google Colab runs notebooks on Google hardware, but free sessions time out and GPU access is not guaranteed. Datalore includes 120 hours of runtime a month with 10 GB of storage, and Deepnote 3 editors on machines with 5 GB of RAM and 2 vCPU. Binder launches any Git repository as a temporary notebook without signup, which suits demos, but nothing persists after the session.

For training and tracking, Weights & Biases and CometML are free for personal use, with W&B including 100 GB of storage. paperspace offers basic instances for public projects only, and Modal gives a $30 monthly credit for Python jobs on CPUs and GPUs. Hex covers 5 projects of SQL and Python notebooks, and PrefectCloud runs 5 deployed data workflows with 500 compute minutes a month.

16 services

Datalore

Hosted Python notebooks from JetBrains with coding assistance and sharing.

120 hrs runtime/mo

10 GB storage

Visit
Deepnote

Collaborative Jupyter-compatible notebooks hosted in the cloud.

3 editors

Unlimited projects, 5 GB RAM, 2 vCPU

Visit
Hex

Collaborative workspace for SQL and Python notebooks, data apps and shared knowledge.

5 projects

Visit
Hugging Face

Hub for sharing ML models and datasets, with hosted inference and training tools.

30,000 input chars/mo

Visit
Kaggle

Data science notebooks with CPU, GPU and TPU runtimes alongside public datasets and competitions.

30 GPU hrs/wk

20 TPU hrs/wk, 4 CPU cores, 30 GB RAM

Visit
paperspace

Cloud GPU platform for building, training and deploying AI models.

5 GB storage

Basic instances, Public projects only

Visit
plot.ly

Online charting for plotting datasets and sharing interactive graphs.

Unlimited public files

10 private files

Visit
PrefectCloud

Hosted orchestration for data pipelines and workflows with serverless execution.

5 deployed workflows

500 compute minutes/mo

Visit

MLOps platform for tracking experiments, versioning datasets and managing models.

  • Personal use

100 GB storage

Visit
B
Binder

Public service that launches a Git repository as runnable interactive notebooks.

  • No sign-up

Limits not listed

Visit
CoCalc

Collaborative cloud workspace with an Ubuntu environment, Jupyter, LaTeX, SageMath and Python.

Limits not listed

Visit
CometML

Experiment tracking, model registry and lineage platform for ML teams.

  • Personal use

Limits not listed

Visit
Google Colab

Hosted Jupyter notebooks that run in the browser on Google-provided compute.

Limits not listed

Visit
RunMat

Browser IDE that runs MATLAB-style .m code with GPU acceleration through WebAssembly and WebGPU.

  • No sign-up
  • Open source

Limits not listed

Visit
Wolfram Language

Cloud access to the Wolfram Language and its curated computational knowledge.

Limits not listed

Visit
Esc

Type to search every entry. ↑↓ to move, Enter to open.