Machine language, also known as machine code, is a computer language that is directly understandable by a computer's CPU (central processing unit), and it is the language into which all programs must be converted before they can be run. Each CPU type has its own machine language, although they are basically fairly similar.

After the source code for a program has been written by one or more humans in a programming language (e.g., C or C++), it is compiled (i.e., converted) into machine code by a specialized program called a compiler, or by an assembler in the case of assembly language. This machine code is then stored as an executable file (i.e., a ready-to-run program) and can be executed (i.e., run) by the operating system any time it is instructed to do so by another program or by a user.

Machine code is extremely difficult for humans to read because it consists merely of patterns of bits (i.e., zeros and ones). Thus, programmers who want to work at the machine code level instead usually use assembly language, which is a human-readable notation for the machine language in which the instructions represented by patterns of zeros and ones are replaced with alphanumeric symbols (called mnemonics) in order to make it easier to remember and work with them (including reducing the chances of making errors). In contrast to high-level languages (e.g., C, C++, Java, Perl and Python), there is a nearly one to one correspondence between a simple assembly language and its corresponding machine language.

Programs for the first electronic computers were written directly in machine code. However, the development of assembly language from the 1950s led to a large increase in programmer productivity. Initially, programs written in assembly language programs were hand-translated into machine code, but this tedious task was later eliminated by the development of assemblers to automate the translations.