NewIntroducing FREE GPU and CPU-backed Jupyter notebooks!Signup for a free account to get started!
As seen on

Effortless infrastructure for Machine Learning and Data Science

1-Click Jupyter Notebooks
Explore, collaborate, and share code and data.
Job runner
Submit tasks to the Paperspace GPU cloud.
Python module
One line of code to run your code on a GPU in the cloud.

Evolve your workflow

We’ve built the ultimate machine learning platform with all the industry standard tools and frameworks like Tensorflow, Torch, Caffe, and Keras pre-installed. Access from anywhere using our web-based terminal, Linux desktop, or SSH.

Caffe machine learning framworkKeras machine learning framworkTensorFlow machine learning framworkTorch machine learning framwork


Launch a GPU-backed VM in less than one minute. Manage everything through a simple, intuitive interface.


The newest and most advanced NVIDIA Pascal GPUs. Fire up multiple machines in parallel to accelerate your workflow.


The most affordable GPUs in the cloud. Paperspace bills by the second so you only pay for what your use.


Multiple layers of security and protection means that only you have access to your data.

Full Linux desktop in a web browser
SSH or web terminal

More Features

  • Jupyter pre-installed
  • Snapshots and Templates
  • Sharing and collaboration tools
  • Pre-configured templates
  • Public IP addresses
  • Desktop or Terminal in a web browser
  • 1Gbps fiber internet
  • Drag-and-drop files into your VM
  • Leverage public datasets

A complete community

Paperspace is building a complete community around machine learning and data science -- and providing the tools to get going fast. Tutorials, templates, and more to get you started

Adversarial Autoencoders (with Pytorch)

Create art using deep neural networks

Gentle Introduction to Data Classification with R

Intro to SciKit Learn

ML Framework comparison

Generative Adverserial Networks (GAN)

Coming Soon

Jupyter Notebooks in the cloud

Building AlexNet

Coming Soon

Generative Adverserial Networks (GAN)

Coming Soon

Adverserial Examples - Fooling a neural network

Coming Soon

Building AlexNet

Coming Soon


We run the latest Nvidia cards at a lower price than any competitor. Easily scale up or down as your needs change.

1 Some regions have limited GPU availability

Maxwell architecture
1664 CUDA cores
192 GB/s
8GB memory
$0.51/ hr

Pascal architecture
2560 CUDA cores
288 GB/s
16GB memory
$0.78/ hr

Pascal architecture
3840 CUDA cores
432 GB/s
24GB memory
$1.10/ hr

2x the performance for a fraction of the price

Modern GPUs built for machine learning. Paperspace is more powerful and less expensive than AWS, GCP, and Azure

InstanceCUDA coresTeraFLOPSVRAM$ / hr1TB transferGeneration500 GB Storage
AWS g2.xlarge15360.724GB$0.65$9 extraKepler
circa 2012
3.6x faster
+ 200%
$0.51free unlimited bandwidthMaxwell$25/mo
AWS p2.xlarge24962.912GB$0.90$9 extraKepler
circa 2012
3x faster
+ 40%
$0.78free unlimited bandwidthPascal$25/mo
384012.024GB$1.10free unlimited bandwidthPascal$25/mo
5120112*16GB$2.30free unlimited bandwidthVolta
Newest Generation (2017)

*Tensor Performance. Learn more about NVIDIA's Volta linehere.
Join over 100,000 developers on the Paperspace cloud
NYU AI Nexus LabYC AIInsight Data Science AINtopology Element CAD/CAM
Access anywhere
World-class support
Cancel anytime