TerminalBench 4.0

A benchmark that evaluates the ability of large language models to write code within terminal environments.