PyTorch

PyTorch is the dominant open-source machine learning framework, originally developed by Meta's AI Research lab (FAIR) and now governed by the PyTorch Foundation under the Linux Foundation. It provides the core tensor computation and automatic differentiation libraries that researchers and engineers use to build, train, and deploy neural networks—including the foundation models that power the agentic economy.

PyTorch's rise to dominance over Google's TensorFlow represents one of the most significant platform shifts in AI history. Its Pythonic design philosophy and eager execution mode made it the preferred framework for research, and that research dominance eventually translated into production dominance as well. Today, the vast majority of frontier model development—at OpenAI, Anthropic, Meta, and others—uses PyTorch.

In the Seven Layers of the Agentic Economy, PyTorch sits at Layer 5: Data & Knowledge, in the ML Training Frameworks subcategory. It is the compiler of the AI era—translating mathematical intent into the parallel computations that produce intelligence.

Although Meta created and sponsors PyTorch, it has such a prominent independent brand identity and ecosystem that it warrants recognition as a distinct entity in the agentic economy landscape.