It is based on a 32-bit ARM Cortex-M4F CPU, and extends their 16-bit MSP430 line, with a larger address space for code and data, and faster integer and floating point calculation than the MSP430.
The MSP430's 16-bit architecture was already once extended to 20 bits (MSP430X) to accommodate those needs, but the resulting 1 MB limit is still too small, and the instruction set extensions slow down the code execution.
[3] The ARM Cortex-M4F architecture used in the MSP432 line allows up to 4 GB of unified program/data/peripheral memory, and has a built-in single precision IEEE754-compatible Floating Point Unit.
The MSP432 is slightly slower, cheaper and uses significantly less power, and tends to have less of the sophisticated peripherals such as wide 32/64-bit timer units, or the quadrature encoder blocks.
This is a higher cost development board with a 100-pin LQFP ZIF socket used by initial MSP432 chips, and a JTAG and Spy Bi-Wire debug interfaces.