A model using standard 32-bit floating-point numbers to represent weights, providing maximum accuracy but requiring more memory.