PyTorch

PyTorch is an open-source deep learning framework: a Python library for defining neural networks, computing gradients automatically (autograd), and training models on GPUs and other accelerators. Its development began in 2016 through work by the Torch community, with major contributions from Meta and other collaborators; the reference paper by Adam Paszke and colleagues was presented at NeurIPS 2019.

Since September 2022, PyTorch's business governance has sat with the PyTorch Foundation, a Linux Foundation project; technical decisions remain with its maintainers, while Meta continues as a member and contributor. As of July 2026, the latest stable release is PyTorch 2.13, published on July 8 according to the official GitHub repository.

What sets it apart

Its defining trait is eager execution: a model is ordinary Python code that runs line by line, so debugging a neural network feels like debugging any other program. On top of that, autograd records tensor operations and computes gradients dynamically, even through the language's native loops and conditionals. The NeurIPS 2019 paper sums up the philosophy: an imperative style without giving up performance. That combination explains how deeply rooted it is in research, where iterating fast matters as much as running fast.

From PyTorch 2 to today: compilation and deployment

March 2023 brought PyTorch 2.0 and torch.compile, a function that compiles a model to speed it up without changing its code, while staying backward compatible. In the tests published for PyTorch 2.0, torch.compile worked on 93% of 163 open-source models and, on an NVIDIA A100 GPU, sped up training by 43% on average: 21% with 32-bit precision and 51% with automatic mixed precision. For deployment, torch.export traces a model's tensor computation into a graph that can be run with other inputs or serialized, provided the constraints recorded during capture continue to hold. ExecuTorch, the runtime for inference on mobile and embedded devices, reached version 1.0 on October 22, 2025, and became part of PyTorch Core on April 7, 2026, three months before PyTorch 2.13. For interoperability with other inference engines, the ONNX exporter remains available.

Ecosystem and governance

Around the core sit official libraries such as torchvision (images) and torchaudio (audio), and much of the industry builds on top of it: Hugging Face's Transformers library dropped TensorFlow and JAX in version 5 to focus on PyTorch, according to its own migration guide. On governance, PyTorch moved from Meta to the Linux Foundation in September 2022, with AMD, AWS, Google Cloud, Meta, Microsoft, and NVIDIA as founding members of the PyTorch Foundation. On May 7, 2025, the foundation announced its expansion as an umbrella organization and welcomed vLLM, focused on model inference, and DeepSpeed, focused on distributed training. The 2022 announcement reported more than 2,400 contributors and nearly 154,000 projects built with PyTorch, not 18,000 user organizations. We give no current usage share because the open official sources reviewed publish no comparable measurement.

When to choose it

PyTorch is the default choice if you start from models published by the research community or the Hugging Face ecosystem, if you want to debug in plain Python, or if your final target is a mobile device via ExecuTorch. Alternatives such as JAX offer advantages in certain functional workflows and on TPUs, and TensorFlow remains present in systems already in production; but for starting a deep learning project today, PyTorch is the entry point with the strongest backing in documentation, tooling and community.