NumPy is the fundamental Python package for scientific computing, providing efficient multidimensional arrays, matrices, and mathematical operations. Its array model is the foundation for much of Python's numerical and data-science ecosystem.
Python's built-in lists can hold numbers, but NumPy arrays are designed for dense numerical data and support vectorized operations, so a calculation can apply to many values without an explicit Python loop. NumPy also supplies mathematical functions, broadcasting, indexing tools, and interfaces to optimized low-level code.
The project grew from earlier numerical packages, especially Numeric and its successor Numarray. Travis Oliphant unified important ideas from those projects and released NumPy in 2006. SciPy builds on NumPy for higher-level scientific algorithms, while Matplotlib focuses on visualization and Pandas provides labeled tabular data structures. Those libraries are closely related, but they are not the fundamental array package named by the question.