PyTorch

What it does

PyTorch is an open source machine learning framework for defining, training, and deploying neural network models in Python. It supports transitioning between eager and graph modes with TorchScript, serving models with TorchServe, and scalable distributed training through the torch.distributed backend. Install options include pip, LibTorch for C++, and source builds, with CUDA, ROCm, and CPU compute platforms.

People use PyTorch to build computer vision, NLP, audio, recommender, and reinforcement learning systems, supported by domain libraries such as torchvision, torchaudio, torchtune for fine-tuning large language models, torchrec, and torchrl. It can be installed locally or launched on cloud platforms including AWS, Google Cloud, Microsoft Azure, Lightning Studios, and Alibaba Cloud. ExecuTorch, a related project, targets on-device inference across mobile and edge hardware.

PyTorch suits machine learning researchers, developers, and teams moving models into production. The latest stable release requires Python 3.10 or later, and prerequisites such as numpy depend on the chosen package manager. Preview builds are generated nightly and are not fully tested or supported. Cloud partner access and billing are handled by those platforms.

What it can do

• Eager and graph execution modes with TorchScript for production paths
• Distributed training through the torch.distributed backend
• Model serving with TorchServe
• Domain libraries for computer vision, NLP, audio, recommender systems, and reinforcement learning
• Fine-tuning of large language models via torchtune
• On-device inference across mobile and edge devices through ExecuTorch

Works with

• AWS
• Google Cloud Platform
• Microsoft Azure
• Lightning Studios
• Alibaba Cloud PAI

Before you choose

• The site notes that Preview builds are nightly and not fully tested or supported, so Stable is suited to most users.
• LibTorch is available only for C++ according to the install page.
• For hardware not listed on the install page, the site points to an Additional Platforms page.

Listed under

• Coding
• Sandboxes and infrastructure
• Deep learning
• Developer framework
• Machine learning
• Open source
• Python
• Research

Get started

Select OS, package, language, and compute platform on the install page.

1. Choose your OS, package manager, language, and compute platform on the install page.
2. Confirm prerequisites such as numpy and Python 3.10 or later, then run the displayed install command.
3. Use the docs and tutorials to verify the installation and begin model work.

Cloud platform usage is billed by those providers; check each provider for current rates.

We couldn’t find published prices for PyTorch.

Sources for this overview

Product pages, documentation, and pricing checked 29 Oct 2026. Each source supports part of the overview; availability and prices can change.

• PyTorch Foundation - PyTorch ↗pytorch.org
• Domains - PyTorch ↗pytorch.org
• Executorch - PyTorch ↗pytorch.org

Company website

pytorch.org: product, pricing and policy pages

YC directory mirror

Batch and company name

Social accounts

Accounts the company links, and its posts on X

Human review

Checked by a person

knowledge_base

manual_review

Last checked 29 Oct 2026. Every field keeps its source and date; unknown means we couldn’t confirm it.