C++ extends C with object-oriented features, templates for generic programming, and a much larger standard library, while keeping the low-level memory control and compiled performance that made C useful for systems programming. It supports multiple programming paradigms within a single codebase: procedural, object-oriented, generic, and to a lesser extent functional, which makes it flexible but also notoriously complex, with a specification and feature set that has grown substantially across each major revision since C++11. Like C, it requires manual memory management unless you use smart pointers and RAII patterns, which the modern standard library encourages but doesn't enforce.
C++ remains the default choice for software where both raw performance and reasonably high-level abstractions matter: game engines, browsers, real-time audio and video processing, machine learning training frameworks, and databases. It's particularly common in codebases that started decades ago and have accumulated large amounts of performance-critical, hard-to-replace logic, as well as new projects where nothing else offers the same combination of speed and control with a mature ecosystem of libraries and tooling.
Repos on this page with substantial C++ codebases include PyTorch, whose performance-critical tensor operations and GPU kernels are implemented in C++ beneath the Python API, OBS Studio, which relies on C++ for real-time video capture and encoding, and vLLM, which uses C++ for its inference engine internals. React Native and Expo also use C++ in their bridge layers to communicate efficiently between JavaScript and native platform code.